chore: remove orphaned SPEC.md and TASKS.md

Both files lived at the repo root as pre-plugin-layout artifacts. On
current main neither is referenced from README, SKILL.md, AGENTS.md,
CHANGELOG, or docs/ — no inbound links to break by removing. Git history
preserves the content for anyone who needs to dig it up.

Closes #352, #353. The PRs by @dinakars777 correctly flagged the drift;
deletion is the cleaner resolution than annotating them as historical.

.claude-plugin/marketplace.json

@@ -11,7 +11,7 @@
     {
       "name": "last30days",
       "description": "Research any topic across Reddit, X, YouTube, TikTok, Instagram, Hacker News, Polymarket, GitHub, and 5+ more sources. AI agent scores by upvotes, likes, and real money - not editors.",
-      "version": "3.2.3",
+      "version": "3.2.4",
       "author": {
         "name": "Matt Van Horn",
         "url": "https://github.com/mvanhorn"

.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "last30days",
-  "version": "3.2.3",
+  "version": "3.2.4",
   "description": "Research any topic across Reddit, X, YouTube, TikTok, Instagram, Hacker News, Polymarket, GitHub, and 5+ more sources. AI agent scores by upvotes, likes, and real money - not editors.",
   "author": {
     "name": "Matt Van Horn",

.github/workflows/security.yml

@@ -0,0 +1,67 @@
+name: Security
+
+on:
+  pull_request:
+  push:
+    branches:
+      - main
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  dependency-audit:
+    name: Dependency audit
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v5
+
+      - name: Set up Python
+        run: uv python install 3.12
+
+      - name: Export locked dependency set
+        run: |
+          uv export \
+            --locked \
+            --all-groups \
+            --no-hashes \
+            --format requirements.txt \
+            --output-file /tmp/last30days-requirements.txt
+
+      # Advisory-first: visibility before enforcement. This repo handles API keys,
+      # cookies, browser tokens, and local env files, so dependency CVEs should be
+      # visible in CI logs even before the project has a clean blocking baseline.
+      # Set continue-on-error: false once a clean baseline run is confirmed.
+      - name: Run pip-audit against locked dependencies
+        continue-on-error: true
+        run: uvx --python 3.12 pip-audit -r /tmp/last30days-requirements.txt --progress-spinner=off
+
+  secret-scan:
+    name: Secret scan
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout full history for diff-aware scanning
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      # Advisory-first: this reports verified secrets in pull requests and pushes to
+      # main, but does not block merges until maintainers confirm a clean baseline.
+      # The TruffleHog action automatically scans the PR range for pull_request
+      # events and the pushed commit range for push events.
+      # Set continue-on-error: false once a clean baseline run is confirmed.
+      # Contributor policy: never commit real secrets in fixtures, tests, docs, or
+      # examples; use obvious dummy values and env-based auth patterns instead.
+      - name: Run TruffleHog OSS secret scan
+        if: github.event_name == 'pull_request' || github.event_name == 'push' || github.event_name == 'workflow_dispatch'
+        uses: trufflesecurity/trufflehog@v3.95.2
+        continue-on-error: true
+        with:
+          path: ./
+          version: v3.95.2
+          extra_args: --only-verified

.github/workflows/validate.yml

@@ -10,7 +10,7 @@ permissions:
   contents: read
 
 jobs:
-  plugin-contract:
+  tests:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
@@ -22,5 +22,5 @@ jobs:
       - name: Set up Python
         run: uv python install 3.12
 
-      - name: Run plugin contract tests
+      - name: Run test suite
-        run: uv run pytest tests/test_plugin_contract.py tests/test_version_consistency.py
+        run: uv run pytest

AGENTS.md

@@ -1 +1,38 @@
-@CLAUDE.md
+# last30days Skill
+
+Agent Skills package for researching any topic across Reddit, X, YouTube, and web. Installable across Claude Code (most common host), Codex, Cursor, GitHub Copilot, Gemini CLI, and 50+ other [Agent Skills](https://agentskills.io) hosts. Python scripts with multi-source search aggregation.
+
+## Structure
+- `skills/last30days/SKILL.md` — canonical skill definition
+- `skills/last30days/scripts/last30days.py` — main research engine
+- `skills/last30days/scripts/lib/` — search, enrichment, rendering modules
+- `skills/last30days/scripts/lib/vendor/bird-search/` — vendored X search client
+- `docs/solutions/` — documented solutions to past problems (bugs, best practices, workflow patterns), organized by category with YAML frontmatter (`module`, `tags`, `problem_type`)
+- `CONCEPTS.md` — shared domain vocabulary (Skill, Engine, Harness, Beta channel) — relevant when orienting to the codebase or discussing project terminology
+
+## Orientation
+- This is an Agent Skills package, not a CLI tool. The product is the slash-command-invoked skill (`/last30days <topic>` in most harnesses); `scripts/last30days.py` is implementation. Claude Code is the most common host but not the only one — features must work across every harness the skill installs into.
+- Feature design starts from the slash-command UX. A new engine flag with no SKILL.md integration is incomplete — the model invoking the skill won't know the flag exists.
+- README and PR examples show `/last30days <topic>` first. Direct CLI invocation (`python3 scripts/last30days.py ...`) is a fallback for scripting, cron, and dev-time engine testing; label it as such, never as the primary path.
+- Slash commands don't pass shell mechanics through. `/last30days OpenClaw --emit=html | pbcopy` is invalid in any harness — either use the slash form (no flags or pipes; let the model translate user intent into engine flags) or use the direct CLI form (full `python3 ...` with explicit flags and a real shell).
+
+## Commands
+```bash
+# Dev/fallback: direct engine invocation (scripting, cron, or engine testing only)
+python3 skills/last30days/scripts/last30days.py "test query" --emit=compact
+npx skills add . -g -y   # one-time: symlink this repo into every detected harness's skill dir
+
+## Rules
+- `lib/__init__.py` must be bare package marker (comment only, NO eager imports)
+- One-time setup: `npx skills add . -g -y` creates symlinks from each detected harness's skill dir to this repo. Edits in the working tree propagate live to every harness — no re-deploy step needed.
+- Git remote: origin = public (`mvanhorn/last30days-skill`)
+
+## Security hygiene
+- Never commit real API keys, browser cookies, auth tokens, app passwords, access tokens, or `.env` contents.
+- Use the env-based auth patterns in `skills/last30days/scripts/lib/env.py`; tests and fixtures must use obvious dummy values only.
+- Keep examples safe by redacting secrets and avoiding copy/pasteable live credentials in docs, fixtures, and test data.
+- Do not weaken or disable the advisory security workflow (`.github/workflows/security.yml`) without explaining why in the PR description or review thread.
+
+## Beta channel
+
+Experimental changes get tested on `mvanhorn/last30days-skill-private`, which installs as a parallel `/last30days-beta` slash command. Beta-only changes never ship to public without a review PR here. Workflow guide lives at `BETA.md` in the private repo. Plan that established this setup: `docs/plans/2026-04-17-005-feat-beta-skill-from-private-repo-plan.md`.

CHANGELOG.md

@@ -7,8 +7,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+
+- `LAST30DAYS_YOUTUBE_SSH_HOST` env var: when set, yt-dlp YouTube search invocations are routed through `ssh <host>` for residential-IP egress. Bypasses YouTube's bot-wall on datacenter IPs (Hetzner/DigitalOcean/AWS) where `ytsearch:` returns 0 results regardless of cookies (the IP fingerprint is checked first). The named host must be configured in `~/.ssh/config` and have yt-dlp installed. Host value is validated against `^[a-zA-Z0-9._-]+$` to reject SSH option-injection (e.g. a leading `-` masquerading as a flag). The transcript path is unchanged (uses the existing HTTP fallback when SSH-routing is on, since the timedtext API isn't bot-walled).
+
 ### Changed
 
+- Replace the SKILL_ROOT resolver loops in Step 1 and comparison-mode with a single `SKILL_DIR` substitution pattern. The model templates the absolute path of the SKILL.md's own directory (which it always knows from the Read tool result); the bash block just validates that `scripts/last30days.py` lives there. Removes ~80 lines of bash across the two locations. Fixes a real bug: the previous resolver could pick a different install than the SKILL.md the model loaded from (spec-vs-engine divergence) and didn't enumerate harnesses like Hermes at all. The simplification works for any harness without enumeration because it just uses wherever SKILL.md was loaded from. STEP 0's marketplaces-stale-clone hop is unchanged.
 - Rename "Digg AI 1000" to just "Digg" in user-facing output (footer line, source label, inline-quote suffix, why_relevant, container attribution). Internal references to the upstream Digg AI 1000 product remain in code comments and docstrings.
 - Bump `POSTS_PER_CLUSTER` from 3 to 5 and the render-side display limit from 2 to 3 to match the per-source enrichment caps used by Reddit, HN, YouTube, TikTok, and GitHub. The previous 3/2 caps routinely truncated cluster context (e.g. dropped a Jason Calacanis quote tweet on a `cli-printing-press` run).
 - Rewrite SKILL.md path resolution. STEP 0 narrows from a global canonical-path enforcement to a Claude-Code-marketplaces-only stale-clone guard. Step 1 SKILL_ROOT resolver walks a single precedence list (Claude plugin cache, then `~/.codex/skills/`, `~/.agents/skills/`, repo checkout, `./.skills/last30days` for `npx skills add`, CWD, Gemini). Adds SKILL.md frontmatter fallback to `render.py::_skill_version` so the badge no longer prints `v?` on installs that don't include `.claude-plugin/plugin.json`.

CLAUDE.md

@@ -1,25 +1 @@
-# last30days Skill
+@AGENTS.md
-
-Claude Code skill for researching any topic across Reddit, X, YouTube, and web.
-Python scripts with multi-source search aggregation.
-
-## Structure
-- `skills/last30days/SKILL.md` — canonical skill definition
-- `skills/last30days/scripts/last30days.py` — main research engine
-- `skills/last30days/scripts/lib/` — search, enrichment, rendering modules
-- `skills/last30days/scripts/lib/vendor/bird-search/` — vendored X search client
-
-## Commands
-```bash
-python3 skills/last30days/scripts/last30days.py "test query" --emit=compact
-npx skills add . -g -y   # one-time: symlink this repo into every detected harness's skill dir
-```
-
-## Rules
-- `lib/__init__.py` must be bare package marker (comment only, NO eager imports)
-- One-time setup: `npx skills add . -g -y` creates symlinks from each detected harness's skill dir to this repo. Edits in the working tree propagate live to every harness — no re-deploy step needed.
-- Git remote: origin = public (`mvanhorn/last30days-skill`)
-
-## Beta channel
-
-Experimental changes get tested on `mvanhorn/last30days-skill-private`, which installs as a parallel `/last30days-beta` slash command. Beta-only changes never ship to public without a review PR here. Workflow guide lives at `BETA.md` in the private repo. Plan that established this setup: `docs/plans/2026-04-17-005-feat-beta-skill-from-private-repo-plan.md`.

CONCEPTS.md

@@ -0,0 +1,23 @@
+# Concepts
+
+Shared vocabulary for `last30days-skill`. Terms here have a precise project-specific meaning — distinct enough from their general technical sense that a new contributor would need them defined to follow conversations, PR descriptions, or the SKILL.md contract.
+
+## The package
+
+### Skill
+
+A self-contained agent-instructions package consisting of a `SKILL.md` prose contract plus a sibling `scripts/` directory containing the executable code the SKILL.md invokes. The package conforms to the [Agent Skills](https://agentskills.io) open format and installs across every major harness (Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and 50+ others) via `npx skills add`, harness-native plugin installers, or per-harness skill directories. A Skill is the unit of distribution; the Skill is the product.
+
+### Engine
+
+The Python script (`scripts/last30days.py`) the Skill's SKILL.md invokes to do the actual research work. The Engine and SKILL.md have a contract: SKILL.md tells the model which flags to pass (`--plan`, `--competitors-plan`, `--x-handle`, `--subreddits`, `--emit=compact`, etc.), and the Engine produces a specific output shape (badge line, ranked evidence clusters, emoji-tree footer) that the model is contractually required to pass through. The Engine is implementation; the SKILL.md prose is the agent-facing surface.
+
+### Harness
+
+The agent runtime that loads Skills and invokes them on the user's behalf. Claude Code is the most common Harness for this Skill but not the only one — Codex, Cursor, GitHub Copilot, Gemini CLI, and the rest of the Agent Skills ecosystem also count. "Multi-harness" describes a Skill that works correctly across every Harness it installs into; features written without multi-harness awareness (e.g., engine flags with no SKILL.md integration, or paths hardcoded to one Harness's install layout) regress on Harnesses other than the one they were tested against.
+
+## Distribution
+
+### Beta channel
+
+A parallel install of the Skill, sourced from the private `mvanhorn/last30days-skill-private` repo and installed as `/last30days-beta` rather than `/last30days`. The Beta channel exists so experimental changes can be tested by real users before they ship to the public `/last30days`. Promotion from Beta to public happens via a review PR against this (public) repo — Beta-only changes never ship to public without that PR. The Beta channel workflow guide lives in `BETA.md` in the private repo.

CONTRIBUTORS.md

@@ -23,7 +23,7 @@ v3 has full GitHub search: issues, PRs, person-mode profiles, project-mode repos
 ### @thinkun
 [PR #116](https://github.com/mvanhorn/last30days-skill/pull/116) - Resilient Reddit, prevent enrichment timeout from discarding results
 v3 has parallel enrichment with per-item timeouts. No results are ever dropped.
-> _Add your bio, website, or anything you'd like here._
+> Thinker, technologist, AI expert, music-tinkerer. Founder of [Thinkun](https://thinkun.com). [@thinkun on GitHub](https://github.com/thinkun) · [@unthink on X](https://x.com/unthink)
 
 ### @thomasmktong
 [PR #124](https://github.com/mvanhorn/last30days-skill/pull/124) - Pure Python Reddit fallback

HERMES_SETUP.md

@@ -51,7 +51,7 @@ On first run, the skill will guide you through setup:
 
 2. **Optional: ScrapeCreators**
    - Adds TikTok, Instagram, Reddit backup
-   - 10,000 free API calls
+   - 100 free credits (no expiration)
    - Sign up at scrapecreators.com
 
 3. **Optional: API Keys**

README.md

@@ -152,8 +152,10 @@ Say "eli5 on" after any research run. The synthesis rewrites in plain language.
 
 - **Free Reddit comments.** Public JSON gives you threads + top comments with upvote counts. No API key, no ScrapeCreators. Just works.
 - **YouTube transcripts that actually work.** Widened candidate pool 3x past music videos to reach talk/review content with captions.
-- **Threads, Pinterest, YouTube + TikTok comments.** Opt-in sources via ScrapeCreators. Set `INCLUDE_SOURCES=tiktok,instagram` and add threads, pinterest, youtube_comments, tiktok_comments for more. `youtube_comments` and `tiktok_comments` surface top comments with vote counts the same way Reddit does.
+- **TikTok, Instagram, Threads.** All three activate automatically once `SCRAPECREATORS_API_KEY` is set — same key, same per-call cost. Suppress any of them with `EXCLUDE_SOURCES=tiktok,instagram,threads` (any comma-separated subset).
-- **Perplexity Sonar.** Grounded web search with citations via OpenRouter. Add `OPENROUTER_API_KEY` to unlock.
+- **Pinterest.** Per-query opt-in (visual pins, narrow utility): the model passes `--search=pinterest` for the runs that need it. Requires `SCRAPECREATORS_API_KEY`.
+- **YouTube + TikTok comments.** Persistent opt-in via `INCLUDE_SOURCES=youtube_comments,tiktok_comments` because each video pulls N extra ScrapeCreators calls on top of the base search. Surface top comments with vote counts the same way Reddit does.
+- **Perplexity Sonar.** Grounded web search with citations via OpenRouter. Add `OPENROUTER_API_KEY` and `INCLUDE_SOURCES=perplexity` (it's a separate paid API — opt-in keeps you from being surprise-billed).
 - **Polymarket noise filtering.** Common-word disambiguation prevents "Apple" from matching "Will Apple release a car?"
 - **Resilient Reddit.** Timeout budgets and runtime fallback. One slow thread doesn't kill the whole run.
 - **Fun judge v2.** Humor scoring baked into the narrative. Reddit's cleverest one-liners mixed into the synthesis where they fit, not dumped in a separate section.
@@ -256,10 +258,28 @@ These platforms don't have relationships with each other. X doesn't know what Re
 | X / Twitter | Log into x.com in any browser | Free |
 | YouTube | `brew install yt-dlp` | Free |
 | Bluesky | App password from bsky.app | Free |
-| TikTok + Instagram + Threads + Pinterest + YouTube comments | ScrapeCreators key | 10,000 free calls |
+| TikTok + Instagram + Threads + Pinterest + YouTube comments | ScrapeCreators key | 100 free credits, then PAYG |
 | Perplexity Sonar | OpenRouter key | Pay as you go |
 | Web search | Brave Search key | 2,000 free queries/month |
 
+### macOS Keychain (optional)
+
+On macOS you can store keys in the system Keychain instead of a `.env` file. The skill picks them up automatically as the lowest-priority source — `.env` files and process environment still win on collision.
+
+```bash
+# Interactive setup — prompts for each known key, skip with empty input
+skills/last30days/scripts/setup-keychain.sh
+
+# Or store a single key by hand
+security add-generic-password -a "$USER" -s last30days-XAI_API_KEY -w "xai-..."
+
+# Inspect / clean up
+skills/last30days/scripts/setup-keychain.sh --list
+skills/last30days/scripts/setup-keychain.sh --delete XAI_API_KEY
+```
+
+Items are stored under service name `last30days-<KEY>` for the current user. On non-Darwin platforms the loader is a no-op, so there is no behaviour change for Linux/Windows users.
+
 ## How it works
 
 1. **You type a topic.** Person, company, product, technology, "X vs Y." Anything.

SPEC.md

@@ -1,77 +0,0 @@
-# last30days Skill Specification
-
-## Overview
-
-`last30days` is a Claude Code skill that researches a given topic across Reddit and X (Twitter) using the OpenAI Responses API and xAI Responses API respectively. It enforces a strict 30-day recency window, popularity-aware ranking, and produces actionable outputs including best practices, a prompt pack, and a reusable context snippet. OpenAI auth can come from `OPENAI_API_KEY` or Codex login credentials.
-
-The skill operates in three modes depending on available API keys: **reddit-only** (OpenAI key), **x-only** (xAI key), or **both** (full cross-validation). It uses automatic model selection to stay current with the latest models from both providers, with optional pinning for stability.
-
-## Architecture
-
-The orchestrator (`last30days.py`) coordinates discovery, enrichment, normalization, scoring, deduplication, and rendering. Each concern is isolated in `scripts/lib/`:
-
-- **env.py**: Load API keys from `~/.config/last30days/.env` and Codex auth from `~/.codex/auth.json`
-- **dates.py**: Date range calculation and confidence scoring
-- **cache.py**: 24-hour TTL caching keyed by topic + date range
-- **http.py**: stdlib-only HTTP client with retry logic
-- **models.py**: Auto-selection of OpenAI/xAI models with 7-day caching
-- **openai_reddit.py**: OpenAI Responses API + web_search for Reddit
-- **xai_x.py**: xAI Responses API + x_search for X
-- **reddit_enrich.py**: Fetch Reddit thread JSON for real engagement metrics
-- **hackernews.py**: Hacker News search via Algolia API (free, no auth)
-- **polymarket.py**: Polymarket prediction market search via Gamma API (free, no auth)
-- **normalize.py**: Convert raw API responses to canonical schema
-- **score.py**: Compute popularity-aware scores (relevance + recency + engagement)
-- **dedupe.py**: Near-duplicate detection via text similarity
-- **render.py**: Generate markdown and JSON outputs
-- **schema.py**: Type definitions and validation
-
-## Embedding in Other Skills
-
-Other skills can import the research context in several ways:
-
-### Inline Context Injection
-```markdown
-## Recent Research Context
-!python3 ~/.claude/skills/last30days/scripts/last30days.py "your topic" --emit=context
-```
-
-### Read from File
-```markdown
-## Research Context
-!cat ~/.local/share/last30days/out/last30days.context.md
-```
-
-### Get Path for Dynamic Loading
-```bash
-CONTEXT_PATH=$(python3 ~/.claude/skills/last30days/scripts/last30days.py "topic" --emit=path)
-cat "$CONTEXT_PATH"
-```
-
-### JSON for Programmatic Use
-```bash
-python3 ~/.claude/skills/last30days/scripts/last30days.py "topic" --emit=json > research.json
-```
-
-## CLI Reference
-
-```
-python3 ~/.claude/skills/last30days/scripts/last30days.py <topic> [options]
-
-Options:
-  --refresh           Bypass cache and fetch fresh data
-  --mock              Use fixtures instead of real API calls
-  --emit=MODE         Output mode: compact|json|md|context|path (default: compact)
-  --sources=MODE      Source selection: auto|reddit|x|both (default: auto)
-```
-
-## Output Files
-
-All outputs are written to `~/.local/share/last30days/out/`:
-
-- `report.md` - Human-readable full report
-- `report.json` - Normalized data with scores
-- `last30days.context.md` - Compact reusable snippet for other skills
-- `raw_openai.json` - Raw OpenAI API response
-- `raw_xai.json` - Raw xAI API response
-- `raw_reddit_threads_enriched.json` - Enriched Reddit thread data

TASKS.md

@@ -1,47 +0,0 @@
-# last30days Implementation Tasks
-
-## Setup & Configuration
-- [x] Create directory structure
-- [x] Write SPEC.md
-- [x] Write TASKS.md
-- [x] Write SKILL.md with proper frontmatter
-
-## Core Library Modules
-- [x] scripts/lib/env.py - Environment and API key loading
-- [x] scripts/lib/dates.py - Date range and confidence utilities
-- [x] scripts/lib/cache.py - TTL-based caching
-- [x] scripts/lib/http.py - HTTP client with retry
-- [x] scripts/lib/models.py - Auto model selection
-- [x] scripts/lib/schema.py - Data structures
-- [x] scripts/lib/openai_reddit.py - OpenAI Responses API
-- [x] scripts/lib/xai_x.py - xAI Responses API
-- [x] scripts/lib/reddit_enrich.py - Reddit thread JSON fetcher
-- [x] scripts/lib/normalize.py - Schema normalization
-- [x] scripts/lib/score.py - Popularity scoring
-- [x] scripts/lib/dedupe.py - Near-duplicate detection
-- [x] scripts/lib/render.py - Output rendering
-
-## Main Script
-- [x] scripts/last30days.py - CLI orchestrator
-
-## Fixtures
-- [x] fixtures/openai_sample.json
-- [x] fixtures/xai_sample.json
-- [x] fixtures/reddit_thread_sample.json
-- [x] fixtures/models_openai_sample.json
-- [x] fixtures/models_xai_sample.json
-
-## Tests
-- [x] tests/test_dates.py
-- [x] tests/test_cache.py
-- [x] tests/test_models.py
-- [x] tests/test_score.py
-- [x] tests/test_dedupe.py
-- [x] tests/test_normalize.py
-- [x] tests/test_render.py
-
-## Validation
-- [x] Run tests in mock mode
-- [x] Demo --emit=compact
-- [x] Demo --emit=context
-- [x] Verify file tree

docs/plans/2026-04-22-002-feat-competitors-flag-comparison-fanout-plan.md

@@ -1,4 +1,7 @@
 ---
+
+> **NOTE (added 2026-05-16):** This plan references `bash scripts/sync.sh`. That script was deleted in [PR #405](https://github.com/mvanhorn/last30days-skill/pull/405); the install workflow is now `npx skills add . -g -y` (symlinks the working tree across every detected harness). For context on why sync.sh went away, see [docs/solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md](../solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md). The decisions captured in this plan remain accurate; only the deploy mechanism changed.
+
 title: "feat: --competitors flag for auto-discovered comparison fan-out"
 type: feat
 status: active

docs/plans/2026-04-22-003-fix-competitors-per-entity-resolution-plan.md

@@ -1,4 +1,7 @@
 ---
+
+> **NOTE (added 2026-05-16):** This plan references `bash scripts/sync.sh`. That script was deleted in [PR #405](https://github.com/mvanhorn/last30days-skill/pull/405); the install workflow is now `npx skills add . -g -y` (symlinks the working tree across every detected harness). For context on why sync.sh went away, see [docs/solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md](../solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md). The decisions captured in this plan remain accurate; only the deploy mechanism changed.
+
 title: "fix: per-entity resolution, default-2, and stale-path guard for --competitors"
 type: fix
 status: active

docs/plans/2026-04-22-005-feat-vs-mode-n-passes-competitors-auto-discovery-plan.md

@@ -1,4 +1,7 @@
 ---
+
+> **NOTE (added 2026-05-16):** This plan references `bash scripts/sync.sh`. That script was deleted in [PR #405](https://github.com/mvanhorn/last30days-skill/pull/405); the install workflow is now `npx skills add . -g -y` (symlinks the working tree across every detected harness). For context on why sync.sh went away, see [docs/solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md](../solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md). The decisions captured in this plan remain accurate; only the deploy mechanism changed.
+
 title: "feat: vs mode runs N full passes and --competitors is vs with auto-discovery"
 type: feat
 status: active

docs/plans/2026-04-22-006-fix-comparison-title-skill-attribution-plan.md

@@ -1,4 +1,7 @@
 ---
+
+> **NOTE (added 2026-05-16):** This plan references `bash scripts/sync.sh`. That script was deleted in [PR #405](https://github.com/mvanhorn/last30days-skill/pull/405); the install workflow is now `npx skills add . -g -y` (symlinks the working tree across every detected harness). For context on why sync.sh went away, see [docs/solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md](../solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md). The decisions captured in this plan remain accurate; only the deploy mechanism changed.
+
 title: "fix: comparison title says (/Last30Days) instead of (Last 30 Days)"
 type: fix
 status: active

docs/solutions/architecture/search-quality-eval-manual-by-default-2026-05-10.md

@@ -0,0 +1,82 @@
+---
+title: Search-quality eval is manual by default, not a CI gate on every PR
+date: 2026-05-10
+category: docs/solutions/architecture
+module: skills/last30days/scripts/evaluate_search_quality.py
+problem_type: design_decision
+component: ci_policy
+severity: low
+applies_when:
+  - a contributor proposes wiring search-quality eval into PR CI
+  - a change affects retrieval, ranking, grounding, or synthesis quality and a reviewer asks "why aren't we testing this in CI?"
+  - someone is deciding whether a new evaluator-style script belongs in the default CI workflow
+related_components:
+  - search_quality_evaluation
+  - ci_workflow
+  - llm_judging
+tags:
+  - ci-policy
+  - eval
+  - design-decision
+  - cost-vs-signal
+  - non-determinism
+  - manual-gates
+---
+
+# Search-quality eval is manual by default, not a CI gate on every PR
+
+## Context
+
+`skills/last30days/scripts/evaluate_search_quality.py` compares a baseline revision against a candidate revision across a fixed pool of reviewer topics. It produces two flavors of metrics: deterministic overlap (Jaccard, retention) and LLM-judged quality scores. The natural impulse on seeing an evaluator script is to wire it into CI on every PR — "regression catcher, run it automatically." We deliberately don't.
+
+Three properties of this particular evaluator make CI-on-every-PR the wrong default:
+
+1. **Live API access.** The candidate revision typically needs the engine to actually run, which means real ScrapeCreators calls, real reddit fetches, real YouTube searches. CI runs would either need production credentials or a record/replay fixture set that drifts almost immediately as external APIs change shape.
+
+2. **Cost and latency.** A full eval pass runs the pipeline N times across reviewer topics. Multiplied by every PR (including doc-only PRs), the spend is meaningful and the wall-clock pushes CI from ~30s to many minutes.
+
+3. **Non-determinism in the judging path.** The LLM-judged metrics are valuable for review but depend on judge-model behavior on a given day. A flaky eval that fails 1 PR in 20 because the judge re-scored an item differently is a worse CI signal than no eval at all — it teaches contributors to retry rather than read the result.
+
+The deterministic overlap metrics are useful regression signals but they are not the same as user-facing correctness. A change that improves overlap can degrade synthesis quality; a change that drops overlap can be a deliberate improvement. So even the deterministic side isn't safe to auto-fail on.
+
+## Guidance
+
+### 1. Keep search-quality eval available, just not automatic
+
+The script stays runnable by maintainers and contributors. The pattern is:
+
+```bash
+LAST30DAYS_PYTHON=python3.13 \
+  python3 skills/last30days/scripts/evaluate_search_quality.py \
+  --baseline main --candidate HEAD
+```
+
+Reviewers can request a manual eval run when a PR is in the retrieval/ranking/synthesis path and the risk warrants it. Contributors can run it locally before submitting if they want signal upfront.
+
+### 2. Standard PR CI gates remain deterministic and contract-shaped
+
+`pytest` (offline-safe), plugin-contract checks, version-consistency contracts, ruff/lint. Anything that returns the same answer twice for the same input. Quality-of-output assessment lives outside that loop.
+
+### 3. The middle ground is `workflow_dispatch`, not auto-PR-gating
+
+If maintainers want a GitHub-triggered eval that doesn't make every PR pay the live-API cost, the right shape is a manually-dispatched workflow (or a label-triggered one) — not a `pull_request:` workflow that runs unconditionally. That keeps the cost knob in human hands.
+
+### 4. Revisit if the eval can ever be made offline-deterministic
+
+The blocker is the live-API + non-determinism combination. If a future iteration of the script can compute meaningful Jaccard/retention metrics against static fixtures (no live API calls, no LLM judging), the decision flips and it becomes a candidate for default CI. The decision below tracks that condition; revisit when it's met.
+
+## What this means in practice
+
+- Don't merge PRs that wire `evaluate_search_quality.py` into the default `validate.yml` workflow.
+- Do merge PRs that add `workflow_dispatch` triggers or label-gated runs.
+- When reviewing a retrieval/ranking change, request a manual eval if the diff suggests it could regress quality — don't expect CI to catch it.
+
+## Links
+
+- `skills/last30days/scripts/evaluate_search_quality.py` — the evaluator script
+- `docs/search-quality-eval.md` — user-facing usage documentation
+- `.github/workflows/validate.yml` — the default CI workflow (deterministic gates only)
+
+---
+
+*Adapted from a draft ADR proposed by @hnshah in [#374](https://github.com/mvanhorn/last30days-skill/pull/374), restructured into the `docs/solutions/` convention. The original ADR text correctly identified the constraint; this version adds the "why workflow_dispatch is the middle ground" framing and the revisit-condition.*

docs/solutions/workflow-issues/release-consistency-test-cascade-2026-05-16.md

@@ -0,0 +1,219 @@
+---
+title: Release-time consistency tests cause cascade CI failures across all open PRs
+date: 2026-05-16
+category: docs/solutions/workflow-issues
+module: ci-release-engineering
+problem_type: workflow_issue
+component: testing_framework
+severity: high
+applies_when:
+  - a test asserts consistency between two release-time artifacts (e.g., SKILL.md version and a hardcoded pin in a shell script)
+  - one artifact is updated as part of a version bump and the other requires a manual lockstep update
+  - multiple long-lived PRs are open simultaneously against the same base branch
+symptoms:
+  - every open PR's CI fails after a version bump even though the PRs are unrelated to versioning
+  - the failing test references a stale hardcoded value that was not updated alongside the bumped version
+  - PR authors must rebase and manually fix an artifact they did not touch
+root_cause: missing_workflow_step
+resolution_type: code_fix
+related_components:
+  - development_workflow
+  - documentation
+tags:
+  - ci
+  - release-engineering
+  - consistency-test
+  - version-pin
+  - cascade-failure
+  - test-design
+  - workflow
+---
+
+# Release-time consistency tests cause cascade CI failures across all open PRs
+
+## Context
+
+A `tests/test_version_consistency.py::test_sync_cache_path_uses_skill_version` test was added to enforce that the version string embedded in `skills/last30days/scripts/sync.sh` (a hardcoded plugin-cache path segment) matched the version frontmatter in `skills/last30days/SKILL.md`. The intention was sound: the cache path had to stay in lockstep with the skill version or the sync would silently pull stale files.
+
+The test worked as designed until a release shipped. At that point it turned into a cascade-failure machine:
+
+1. A release PR bumps `SKILL.md` version (e.g., 3.2.0 → 3.2.1) **and** bumps the `sync.sh` pin. That PR's CI is green.
+2. The release PR merges to `main`.
+3. Every PR that was open at merge time was branched from pre-release `main`. Those PRs have `SKILL.md` 3.2.1 (inherited via merge-base with `main`) but their branch never touched `sync.sh`.
+4. CI for those PRs runs the consistency test against the new `main` — `SKILL.md` says 3.2.1, `sync.sh` still says 3.2.0 — and fails.
+5. All open PRs are now red simultaneously, with a failure that has nothing to do with their changes.
+
+This affected at least five PRs during the 2026-05-13 to 2026-05-15 window: PR #400 (caught during rebase, required a manual pin bump), PRs #390 and #392 (OpenClaw `SCRAPECREATORS_API_KEY` fix, both stalled for the same stale-pin reason), and at least two others. A follow-up hotfix PR (#397 — `fix(sync): bump cache target to 3.2.1 to match SKILL.md`) was required just to unblock the queue.
+
+The permanent fix was PR #405: delete `sync.sh` entirely (the install workflow made it redundant) and drop `test_sync_cache_path_uses_skill_version`. Once both were gone, no version-consistency cascade was possible.
+
+## Guidance
+
+### 1. Don't write consistency tests that read two files and assert one matches a substring derived from the other
+
+This pattern looks safe but is not:
+
+```python
+def test_sync_cache_path_uses_skill_version(self) -> None:
+    sync_text = (SKILL_ROOT / "scripts" / "sync.sh").read_text(encoding="utf-8")
+    version = _skill_version()          # reads SKILL.md
+    self.assertIn(
+        f'last30days-skill/last30days/{version}"',
+        sync_text,                      # asserts sync.sh contains that string
+    )
+```
+
+It encodes the assumption that both files are always updated together, in the same commit, on the same branch. That assumption breaks the moment two files have independent lifecycle owners — a versioned manifest and a deployment script are archetypal examples.
+
+### 2. If the values genuinely need to stay in sync, derive one from the other at runtime
+
+Remove the hardcoded pin from `sync.sh` and compute it:
+
+```bash
+# sync.sh — derive version from SKILL.md at runtime, no pin to maintain
+SKILL_VERSION=$(grep -m1 '^version:' "$(dirname "$0")/../SKILL.md" \
+    | sed 's/version:[[:space:]]*"\([^"]*\)"/\1/')
+CACHE_PATH="last30days-skill/last30days/${SKILL_VERSION}"
+```
+
+Now there is only one source of truth (`SKILL.md`). The test that asserted they matched becomes vacuous and should be deleted. If `SKILL.md` is wrong, the sync itself will fail loudly — which is better feedback than a CI gate on a different PR.
+
+### 3. If two values must stay independent for legitimate reasons, update them together and make the test self-skip if either source is missing
+
+If separate versioning is genuinely required (e.g., SKILL.md versions for harness consumers, sync.sh versions a private artifact store with its own cadence), update both in the same PR — never staggered — and write the test to self-skip rather than error when either file is absent:
+
+```python
+def test_sync_cache_path_uses_skill_version(self) -> None:
+    sync_sh = SKILL_ROOT / "scripts" / "sync.sh"
+    if not sync_sh.exists():
+        self.skipTest("sync.sh not present; skipping pin consistency check")
+    sync_text = sync_sh.read_text(encoding="utf-8")
+    version = _skill_version()
+    self.assertIn(
+        f'last30days-skill/last30days/{version}"',
+        sync_text,
+    )
+```
+
+Self-skipping means deleting the file is a non-event in CI — no cascading red, no hotfix PR to the queue.
+
+### 4. Run consistency tests against the merge-base diff, not main
+
+If you keep a two-file consistency test, scope it so it only fails when the PR itself modifies one of the two files but not the other. A GitHub Actions step can do this:
+
+```yaml
+- name: Check sync.sh version pin consistency
+  run: |
+    BASE=$(git merge-base HEAD origin/main)
+    SKILL_CHANGED=$(git diff --name-only "$BASE" HEAD | grep -c 'SKILL\.md' || true)
+    SYNC_CHANGED=$(git diff --name-only "$BASE" HEAD | grep -c 'sync\.sh' || true)
+    if [ "$SKILL_CHANGED" -gt 0 ] && [ "$SYNC_CHANGED" -eq 0 ]; then
+      echo "SKILL.md version bumped but sync.sh pin was not updated"
+      exit 1
+    fi
+```
+
+This only fires when your PR touched `SKILL.md` and left `sync.sh` alone — never because a release merged to `main` after you branched.
+
+### 5. Ask whether you actually need this test
+
+If the values are wrong, downstream tooling will fail loudly: the sync will fetch the wrong artifact, installs will break, or the harness will reject the version. A test that exists only to catch a human-bookkeeping error at release time adds cascade-fail risk without offering a meaningfully earlier signal. Weigh that cost before adding any two-file consistency gate.
+
+## Why This Matters
+
+The damage from a stale-pin consistency test is asymmetric. It:
+
+- Fails on every open PR simultaneously the moment a release lands on `main` — not just the PR that forgot to update the pin.
+- Produces a failure message that points at a line in a test file with no obvious relationship to the PR's actual changes.
+- Requires either a hotfix PR (touching a file the failing PRs have no business touching) or a manual rebase of every affected branch.
+- Blocks work that has already been reviewed and approved.
+
+In this repo the effect was measurable: at least five PRs stalled across a two-day window, one hotfix PR was shipped just to unblock the queue, and multiple authors spent time debugging a failure completely unrelated to their changes.
+
+The broader principle is that tests which gate on *bookkeeping consistency between files* impose their maintenance cost on every contributor, every time, even when those contributors did nothing wrong. That cost compounds with team size and release cadence.
+
+## When to Apply
+
+Apply this guidance whenever you find yourself:
+
+- Writing a test that reads two files and asserts that a string in one matches a value derived from the other.
+- Adding a CI step labeled "consistency check," "sync check," or "pin check" where the check compares a hardcoded value against a computed one from a separate file.
+- Working in a repo where a versioned manifest (e.g., `SKILL.md`, `package.json`, `pyproject.toml`) and a deployment artifact (e.g., a shell script, a Dockerfile, a Helm values file) are both maintained by hand.
+- Reviewing a PR that touches only one of two "paired" files and fails a consistency test for the other.
+
+It does *not* apply to tests that read a single source of truth and validate its internal structure (e.g., asserting that `SKILL.md`'s frontmatter version is double-quoted, or that `package.json`'s `version` field is a valid semver string). Those tests have one file and one assertion; they cannot cascade across branches.
+
+## Examples
+
+### Before — the pattern that caused the cascade
+
+Original `tests/test_version_consistency.py` (deleted in commit `9fb19ea`):
+
+```python
+import re
+import unittest
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+SKILL_ROOT = ROOT / "skills" / "last30days"
+
+
+def _skill_version() -> str:
+    text = (SKILL_ROOT / "SKILL.md").read_text(encoding="utf-8")
+    match = re.search(r'^version:\s*"([^"]+)"\s*$', text, re.MULTILINE)
+    if not match:
+        raise AssertionError("SKILL.md version frontmatter not found")
+    return match.group(1)
+
+
+class TestVersionConsistency(unittest.TestCase):
+    def test_sync_cache_path_uses_skill_version(self) -> None:
+        sync_text = (SKILL_ROOT / "scripts" / "sync.sh").read_text(encoding="utf-8")
+        version = _skill_version()          # source 1: SKILL.md frontmatter
+        self.assertIn(                      # assertion: sync.sh must contain
+            f'last30days-skill/last30days/{version}"',
+            sync_text,                      # source 2: hardcoded string in sync.sh
+        )
+```
+
+`sync.sh` contained a line like:
+
+```bash
+PLUGIN_CACHE="$HOME/.cache/last30days-skill/last30days/3.2.0"
+```
+
+When SKILL.md bumped to `3.2.1` in a release PR, `sync.sh` was updated in the same PR and CI stayed green. But every PR branched before that release still had `sync.sh` at `3.2.0`. Their CI failed immediately, with an assertion error pointing at the test, not at the release PR.
+
+### After — what we did: delete both
+
+PR #405 deleted `sync.sh` (the install workflow replaced it) and dropped `test_sync_cache_path_uses_skill_version` in the same change. No consistency gate, no pin to maintain, no cascade possible.
+
+### After — what we could have done instead: derive at runtime
+
+If `sync.sh` had still been needed, the right fix would have been to remove the hardcoded version from the script and derive it from `SKILL.md`:
+
+```bash
+#!/usr/bin/env bash
+# sync.sh — no hardcoded version; reads SKILL.md as single source of truth
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+SKILL_VERSION=$(grep -m1 '^version:' "${SCRIPT_DIR}/../SKILL.md" \
+    | sed 's/version:[[:space:]]*"\([^"]*\)"/\1/')
+
+if [ -z "$SKILL_VERSION" ]; then
+    echo "error: could not parse version from SKILL.md" >&2
+    exit 1
+fi
+
+PLUGIN_CACHE="$HOME/.cache/last30days-skill/last30days/${SKILL_VERSION}"
+# ... rest of sync logic
+```
+
+With this in place, `test_sync_cache_path_uses_skill_version` has no reason to exist — there is nothing to assert. Delete it. If the version parsing breaks, `sync.sh` itself exits non-zero with a clear message.
+
+## Related
+
+- **PR #397** (merged) — `fix(sync): bump cache target to 3.2.1 to match SKILL.md`. The hotfix that unblocked the cascade temporarily by bumping the pin.
+- **PR #400** (merged) — caught the same cascade during rebase; had to bump the pin to clear CI.
+- **PR #390** (closed) and **PR #392** (rebased + merged) — OpenClaw `SCRAPECREATORS_API_KEY` fix; both blocked by the cascade until rebased onto post-#405 main.
+- **PR #405** (merged) — the permanent fix: deleted `sync.sh` + `test_sync_cache_path_uses_skill_version` together.
+- **PR #412** (merged) — adjacent work that consolidated SKILL.md version parsing into `lib/skill_meta.py`, reducing future drift risk by giving the version field one canonical reader.

greptile.json

@@ -0,0 +1,4 @@
+{
+  "triggerOnUpdates": true,
+  "statusCheck": true
+}

hooks/scripts/check-config.sh

@@ -97,7 +97,20 @@ if [[ -n "$HAS_BSKY" ]]; then
   SOURCE_COUNT=$((SOURCE_COUNT + 1))
 fi
 if [[ -n "$HAS_SCRAPECREATORS" ]]; then
-  SOURCE_COUNT=$((SOURCE_COUNT + 3))  # Reddit comments + TikTok + Instagram
+  # Start with Reddit comments + TikTok + Instagram, subtract any in EXCLUDE_SOURCES.
+  # Normalise EXCLUDED (lowercase + collapse whitespace around commas + strip outer
+  # whitespace) so the matching mirrors pipeline.py's .strip().lower() parsing.
+  SC_ADD=3
+  EXCLUDED="${ENV_EXCLUDE_SOURCES:-${EXCLUDE_SOURCES:-}}"
+  EXCLUDED_NORM=$(printf '%s' "$EXCLUDED" | tr '[:upper:]' '[:lower:]' \
+    | sed -E 's/[[:space:]]*,[[:space:]]*/,/g; s/^[[:space:]]+//; s/[[:space:]]+$//')
+  if [[ ",$EXCLUDED_NORM," == *",tiktok,"* ]]; then
+    SC_ADD=$((SC_ADD - 1))
+  fi
+  if [[ ",$EXCLUDED_NORM," == *",instagram,"* ]]; then
+    SC_ADD=$((SC_ADD - 1))
+  fi
+  SOURCE_COUNT=$((SOURCE_COUNT + SC_ADD))
 fi
 
 if [[ -n "$HAS_SCRAPECREATORS" ]]; then
@@ -107,6 +120,6 @@ else
   # Setup done but missing ScrapeCreators — recommend it
   echo "/last30days: Ready — ${SOURCE_COUNT} sources active."
   echo "  Tip: Add ScrapeCreators for Reddit comments + TikTok + Instagram."
-  echo "  10,000 free API calls, no credit card — scrapecreators.com"
+  echo "  100 free credits, no credit card — scrapecreators.com"
   echo "  last30days has no affiliation with any API provider."
 fi

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "last30days-skill"
-version = "3.2.3"
+version = "3.2.4"
 description = "Multi-source last-30-days research skill"
 readme = "README.md"
 requires-python = ">=3.12"
@@ -8,7 +8,7 @@ dependencies = []
 
 [dependency-groups]
 dev = [
-  "pytest>=9,<10",
+  "pytest>=9.0.3,<10",
   "pytest-cov>=7,<8",
 ]
 

skills/last30days/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: last30days
-version: "3.2.3"
+version: "3.2.4"
 description: "Research what people actually say about any topic in the last 30 days. Pulls posts and engagement from Reddit, X, YouTube, TikTok, Hacker News, Polymarket, GitHub, and the web."
 argument-hint: 'last30days nvidia earnings reaction | last30days AI video tools | last30days what users want in react'
 allowed-tools: Bash, Read, Write, AskUserQuestion, WebSearch
@@ -13,9 +13,9 @@ metadata:
   openclaw:
     emoji: "📰"
     requires:
-      env:
+      env: []
-        - SCRAPECREATORS_API_KEY
       optionalEnv:
+        - SCRAPECREATORS_API_KEY
         - OPENAI_API_KEY
         - XAI_API_KEY
         - OPENROUTER_API_KEY
@@ -97,7 +97,7 @@ You are inside the `/last30days` SKILL. This is a specific research tool with a
 
 **How v3.0.7 fixes it:** three structural anchors.
 1. **The MANDATORY first-line badge** (`🌐 last30days v{VERSION} · synced {YYYY-MM-DD}`) at the top of every response is the LAW 2 / LAW 4 enforcement anchor. See "BADGE (MANDATORY, FIRST LINE OF OUTPUT)" in the synthesis section.
-2. **The SKILL_ROOT resolver** in the engine Bash calls walks a precedence list of known install locations and picks the highest-versioned freshest copy, never `~/.openclaw/` or other stale copies.
+2. **The SKILL_DIR substitution** in the engine Bash calls uses the directory of the SKILL.md the model just Read — no resolver list, no precedence walk. Whichever install the harness loaded SKILL.md from is the install whose engine runs. Aligns spec-with-code and works for any harness without enumerating its install path.
 3. **This preface** tells you plainly: do NOT improvise. Follow SKILL.md top to bottom.
 
 If you catch yourself about to write a `##` section header in a GENERAL-query body, a custom title line, a `Sources:` bullet list, a `for dir in ...` path-discovery loop, or a bare `python3 scripts/last30days.py "{TOPIC}"` engine call with no pre-flight flags — stop. Those are the exact failure modes the LAWs and this contract exist to prevent. The 10/10 beta validation from 2026-04-18 and the 0/8 public v3.0.6 regression from the same day had THE SAME MODEL and SIMILAR SKILL.md CONTENT; the delta is the three anchors this release restores. Read SKILL.md top to bottom before emitting your first response.
@@ -114,7 +114,7 @@ These anchors used to live at line 1094 of this file. Three independent Opus 4.7
 🌐 last30days v{VERSION} · synced {YYYY-MM-DD}
 ```
 
-Replace `{VERSION}` with the installed plugin version (`jq -r '.version' "$SKILL_ROOT/../../.claude-plugin/plugin.json" 2>/dev/null || awk '/^version:/{gsub(/"/,"",$2); print $2; exit}' "$SKILL_ROOT/SKILL.md"`) and `{YYYY-MM-DD}` with today's date. No other text on this line. One blank line after, then the synthesis begins.
+Replace `{VERSION}` with the installed plugin version (`jq -r '.version' "$SKILL_DIR/../../.claude-plugin/plugin.json" 2>/dev/null || awk '/^version:/{gsub(/"/,"",$2); print $2; exit}' "$SKILL_DIR/SKILL.md"`) and `{YYYY-MM-DD}` with today's date. No other text on this line. One blank line after, then the synthesis begins.
 
 **Why the badge is MANDATORY:** it is the structural anchor for the canonical output shape. Without it the model drifts into blog-post narrative format with `##` section headers and invented titles, violating LAW 2 and LAW 4. The 2026-04-18 public v3.0.6 0/8 regression produced outputs with section headers like "The headline", "Why he is everywhere", "1. gstack dominates", "The 'Homecoming' peak". Direct cause: this anchor was absent. Do NOT skip the badge. Do NOT describe it. Do NOT paraphrase it. Emit it verbatim as line 1.
 
@@ -243,7 +243,7 @@ If your Bash call to `last30days.py` does NOT include the FULL pre-flight checkl
 
 ---
 
-# last30days v3.2.3: Research Any Topic from the Last 30 Days
+# last30days v3.2.4: Research Any Topic from the Last 30 Days
 
 > **Permissions overview:** Reads public web/platform data and optionally saves research briefings to `LAST30DAYS_MEMORY_DIR` (defaults to `~/Documents/Last30Days`). X/Twitter search uses optional user-provided tokens (AUTH_TOKEN/CT0 env vars). Bluesky search uses optional app password (BSKY_HANDLE/BSKY_APP_PASSWORD env vars - create at bsky.app/settings/app-passwords). All credential usage and data writes are documented in the [Security & Permissions](#security--permissions) section.
 
@@ -330,12 +330,11 @@ Common patterns:
 - If digg-pp-cli is installed (check `which digg-pp-cli`): add Digg
 - If AUTH_TOKEN/CT0 or XAI_API_KEY or FROM_BROWSER is set, or xurl CLI is installed and authenticated: add X
 - If yt-dlp is installed (check `which yt-dlp`): add YouTube
-- If SCRAPECREATORS_API_KEY is set and INCLUDE_SOURCES contains tiktok: add TikTok
+- If SCRAPECREATORS_API_KEY is set: add TikTok, Instagram, Threads (suppress any of these via EXCLUDE_SOURCES)
-- If SCRAPECREATORS_API_KEY is set and INCLUDE_SOURCES contains instagram: add Instagram
+- If SCRAPECREATORS_API_KEY is set and the user explicitly requested pinterest for this query (e.g. via `--search=pinterest`): add Pinterest
-- If SCRAPECREATORS_API_KEY is set and INCLUDE_SOURCES contains threads: add Threads
-- If SCRAPECREATORS_API_KEY is set and INCLUDE_SOURCES contains pinterest: add Pinterest
 - If BSKY_HANDLE and BSKY_APP_PASSWORD are set: add Bluesky
-- If OPENROUTER_API_KEY is set: add Perplexity
+- If OPENROUTER_API_KEY is set and INCLUDE_SOURCES contains perplexity: add Perplexity
+- If EXCLUDE_SOURCES is set (comma-separated, case-insensitive): drop any matching source from the list above before displaying
 
 Then display (use "and more" if 5+ sources, otherwise list all with Oxford comma):
 
@@ -592,27 +591,21 @@ When the user asks "X vs Y" (or "X vs Y vs Z"), the engine fans out N full `pipe
 
 **Invocation:**
 ```bash
-# Comparison mode skips Step 1, so resolve SKILL_ROOT inline here (same precedence
+# SKILL_DIR = absolute path of the directory containing THIS SKILL.md you just Read.
-# walk as Step 1 — keep the two in sync if you edit either).
+# Substitute the actual path below — your harness told you where this file lives via
-SKILL_ROOT=""
+# the Read tool result. Examples:
-CLAUDE_PLUGIN_ROOT="$(find "$HOME/.claude/plugins/cache/last30days-skill/last30days" -mindepth 1 -maxdepth 1 -type d 2>/dev/null | sort -V | tail -1)"
+#   Read ~/.claude/skills/last30days/SKILL.md      → SKILL_DIR=$HOME/.claude/skills/last30days
-if [ -n "$CLAUDE_PLUGIN_ROOT" ]; then
+#   Read ~/.codex/skills/last30days/SKILL.md       → SKILL_DIR=$HOME/.codex/skills/last30days
-  if [ -f "$CLAUDE_PLUGIN_ROOT/skills/last30days/scripts/last30days.py" ]; then
+#   Read ~/.claude/plugins/cache/last30days-skill/last30days/3.2.4/skills/last30days/SKILL.md
-    SKILL_ROOT="$CLAUDE_PLUGIN_ROOT/skills/last30days"
+#     → SKILL_DIR=$HOME/.claude/plugins/cache/last30days-skill/last30days/3.2.4/skills/last30days
-  elif [ -f "$CLAUDE_PLUGIN_ROOT/scripts/last30days.py" ]; then
+# scripts/last30days.py is always a direct child of SKILL_DIR (every install layout
-    SKILL_ROOT="$CLAUDE_PLUGIN_ROOT"
+# packages SKILL.md and scripts/ as siblings).
-  fi
+SKILL_DIR="<absolute path of the directory containing the SKILL.md you Read>"
-fi
+
-if [ -z "$SKILL_ROOT" ] || [ ! -f "$SKILL_ROOT/scripts/last30days.py" ]; then
+if [ ! -f "$SKILL_DIR/scripts/last30days.py" ]; then
-  for dir in \
+  echo "ERROR: scripts/last30days.py not found under SKILL_DIR=$SKILL_DIR" >&2
-      "$HOME/.codex/skills/last30days" \
+  echo "Re-check the directory of the SKILL.md you Read and substitute it as SKILL_DIR above." >&2
-      "$HOME/.agents/skills/last30days" \
+  exit 1
-      "./skills/last30days" \
-      "./.skills/last30days" \
-      "." \
-      "${GEMINI_EXTENSION_DIR:-}"; do
-    [ -n "$dir" ] && [ -f "$dir/scripts/last30days.py" ] && SKILL_ROOT="$dir" && break
-  done
 fi
 
 # Write the per-entity plan to a tmpfile and pass the path to the engine.
@@ -631,7 +624,7 @@ cat > "$COMPETITORS_PLAN_FILE" <<'PLAN_EOF'
 }
 PLAN_EOF
 
-"${LAST30DAYS_PYTHON}" "${SKILL_ROOT}/scripts/last30days.py" "{TOPIC_A} vs {TOPIC_B} vs {TOPIC_C}" \
+"${LAST30DAYS_PYTHON}" "${SKILL_DIR}/scripts/last30days.py" "{TOPIC_A} vs {TOPIC_B} vs {TOPIC_C}" \
   --emit=compact \
   --save-dir="${LAST30DAYS_MEMORY_DIR}" \
   --save-suffix=v3 \
@@ -916,44 +909,24 @@ Store your plan as `QUERY_PLAN_JSON` - you'll pass it to the script in the next
 **IMPORTANT: Include `--x-handle={RESOLVED_HANDLE}` in the command. For comparison mode: Pass `--x-handle={TOPIC_A_HANDLE}` to the first pass, `--x-handle={TOPIC_B_HANDLE}` to the second pass, and both to the head-to-head pass. Also include `--subreddits={RESOLVED_SUBREDDITS}`, `--tiktok-hashtags={RESOLVED_HASHTAGS}`, `--tiktok-creators={RESOLVED_TIKTOK_CREATORS}`, and `--ig-creators={RESOLVED_IG_CREATORS}` from Step 0.55. Omit any flag where the value was not resolved (empty).**
 
 ```bash
-# Resolve SKILL_ROOT by walking a precedence list of known install locations.
+# SKILL_DIR = absolute path of the directory containing THIS SKILL.md you just Read.
-# Claude Code plugin cache wins when present (highest version dir picked on upgrade),
+# Substitute the actual path below — your harness told you where this file lives via
-# then common per-harness skill dirs, then a repo checkout.
+# the Read tool result. Examples:
-SKILL_ROOT=""
+#   Read ~/.claude/skills/last30days/SKILL.md      → SKILL_DIR=$HOME/.claude/skills/last30days
+#   Read ~/.codex/skills/last30days/SKILL.md       → SKILL_DIR=$HOME/.codex/skills/last30days
+#   Read ~/.claude/plugins/cache/last30days-skill/last30days/3.2.4/skills/last30days/SKILL.md
+#     → SKILL_DIR=$HOME/.claude/plugins/cache/last30days-skill/last30days/3.2.4/skills/last30days
+# scripts/last30days.py is always a direct child of SKILL_DIR (every install layout
+# packages SKILL.md and scripts/ as siblings).
+SKILL_DIR="<absolute path of the directory containing the SKILL.md you Read>"
 
-# 1. Claude Code plugin cache (versioned, sort -V picks freshest). Two cache layouts ship in the wild:
+if [ ! -f "$SKILL_DIR/scripts/last30days.py" ]; then
-#    nested ({cache}/{version}/skills/last30days/scripts/...) and flat ({cache}/{version}/scripts/...).
+  echo "ERROR: scripts/last30days.py not found under SKILL_DIR=$SKILL_DIR" >&2
-# `find` (not `ls + glob`) because zsh errors on globs that match nothing, leaking
+  echo "Re-check the directory of the SKILL.md you Read and substitute it as SKILL_DIR above." >&2
-# noisy "no matches found" stderr in Codex/zsh sessions even with 2>/dev/null.
-CLAUDE_PLUGIN_ROOT="$(find "$HOME/.claude/plugins/cache/last30days-skill/last30days" -mindepth 1 -maxdepth 1 -type d 2>/dev/null | sort -V | tail -1)"
-if [ -n "$CLAUDE_PLUGIN_ROOT" ]; then
-  if [ -f "$CLAUDE_PLUGIN_ROOT/skills/last30days/scripts/last30days.py" ]; then
-    SKILL_ROOT="$CLAUDE_PLUGIN_ROOT/skills/last30days"
-  elif [ -f "$CLAUDE_PLUGIN_ROOT/scripts/last30days.py" ]; then
-    SKILL_ROOT="$CLAUDE_PLUGIN_ROOT"
-  fi
-fi
-
-# 2. Common per-harness skill dirs and repo checkout (npx skills, Codex, Agents, Gemini, etc).
-if [ -z "$SKILL_ROOT" ] || [ ! -f "$SKILL_ROOT/scripts/last30days.py" ]; then
-  for dir in \
-      "$HOME/.codex/skills/last30days" \
-      "$HOME/.agents/skills/last30days" \
-      "./skills/last30days" \
-      "./.skills/last30days" \
-      "." \
-      "${GEMINI_EXTENSION_DIR:-}"; do
-    [ -n "$dir" ] && [ -f "$dir/scripts/last30days.py" ] && SKILL_ROOT="$dir" && break
-  done
-fi
-
-if [ -z "${SKILL_ROOT:-}" ] || [ ! -f "$SKILL_ROOT/scripts/last30days.py" ]; then
-  echo "ERROR: Could not find scripts/last30days.py in any known install location" >&2
-  echo "Searched: ~/.claude/plugins/cache/, ~/.codex/skills/, ~/.agents/skills/, ./skills/last30days, ./.skills/last30days, ." >&2
   exit 1
 fi
 
-"${LAST30DAYS_PYTHON}" "${SKILL_ROOT}/scripts/last30days.py" $ARGUMENTS --emit=compact --save-dir="${LAST30DAYS_MEMORY_DIR}" --save-suffix=v3
+"${LAST30DAYS_PYTHON}" "${SKILL_DIR}/scripts/last30days.py" $ARGUMENTS --emit=compact --save-dir="${LAST30DAYS_MEMORY_DIR}" --save-suffix=v3
 ```
 
 **If you ran Steps 0.55 and 0.75 (agent planning), pass the plan via a tmpfile and add the targeting flags:**
@@ -1715,7 +1688,7 @@ Want another prompt? Just tell me what you're creating next.
 - Sends search queries to Algolia HN Search API (`hn.algolia.com`) for Hacker News story and comment discovery (free, no auth)
 - Sends search queries to Polymarket Gamma API (`gamma-api.polymarket.com`) for prediction market discovery (free, no auth)
 - Runs `yt-dlp` locally for YouTube search and transcript extraction (no API key, public data)
-- Sends search queries to ScrapeCreators API (`api.scrapecreators.com`) for TikTok and Instagram search, transcript/caption extraction (PAYG after 10,000 free API calls)
+- Sends search queries to ScrapeCreators API (`api.scrapecreators.com`) for TikTok and Instagram search, transcript/caption extraction (PAYG after 100 free credits)
 - Optionally sends search queries to Brave Search API, Parallel AI API, or OpenRouter API for web search
 - Fetches public Reddit thread data from `reddit.com` for engagement metrics
 - Stores research findings in local SQLite database (watchlist mode only)
@@ -1728,7 +1701,7 @@ Want another prompt? Just tell me what you're creating next.
 - Does not log, cache, or write API keys to output files
 - Does not send data to any endpoint not listed above
 - Hacker News and Polymarket sources are always available (no API key, no binary dependency)
-- TikTok and Instagram sources require SCRAPECREATORS_API_KEY (10,000 free API calls, then PAYG). Reddit uses ScrapeCreators only as a backup when public Reddit is unavailable.
+- TikTok and Instagram sources require SCRAPECREATORS_API_KEY (100 free credits one-time, then PAYG). Reddit uses ScrapeCreators only as a backup when public Reddit is unavailable.
 - Can be invoked autonomously by agents via the Skill tool (runs inline, not forked); pass `--agent` for non-interactive report output
 
 **Bundled scripts:** `scripts/last30days.py` (main research engine), `scripts/lib/` (search, enrichment, rendering modules), `scripts/lib/vendor/bird-search/` (vendored X search client, MIT licensed)

skills/last30days/scripts/evaluate_search_quality.py

@@ -20,6 +20,7 @@ sys.path.insert(0, str(Path(__file__).parent))
 
 from lib import env as envlib
 from lib import schema
+from lib.providers import GEMINI_FLASH_LITE
 
 
 SKILL_ROOT = Path(__file__).resolve().parents[1]
@@ -43,7 +44,7 @@ def _load_default_topics() -> list[tuple[str, str]]:
 
 DEFAULT_TOPICS = _load_default_topics()
 DEFAULT_SEARCH = ""
-DEFAULT_JUDGE_MODEL = "gemini-3.1-flash-lite-preview"
+DEFAULT_JUDGE_MODEL = GEMINI_FLASH_LITE
 GEMINI_API_URL = "https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent?key={api_key}"
 
 

skills/last30days/scripts/last30days.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 # ruff: noqa: E402
-"""last30days v3.0.0 CLI."""
+"""last30days CLI."""
 
 from __future__ import annotations
 
@@ -97,11 +97,13 @@ def save_output(
     save_dir: str,
     suffix: str = "",
     synthesis_md: str | None = None,
+    topic_override: str | None = None,
+    rendered_content: str | None = None,
 ) -> Path:
     from datetime import datetime
     path = Path(save_dir).expanduser().resolve()
     path.mkdir(parents=True, exist_ok=True)
-    slug = slugify(report.topic)
+    slug = slugify(topic_override or report.topic)
     extension = "json" if emit == "json" else "html" if emit == "html" else "md"
     raw_label = "raw-html" if emit == "html" else "raw"
     suffix_part = f"-{suffix}" if suffix else ""
@@ -110,7 +112,9 @@ def save_output(
         out_path = path / f"{slug}-{raw_label}{suffix_part}-{datetime.now().strftime('%Y-%m-%d')}.{extension}"
     # Markdown saves keep the complete debug artifact. JSON and HTML preserve
     # their requested wire format so file extensions match their content.
-    if emit in {"json", "html"}:
+    if rendered_content is not None:
+        content = rendered_content
+    elif emit in {"json", "html"}:
         content = emit_output(report, emit, synthesis_md=synthesis_md)
     else:
         content = render.render_full(report)
@@ -171,6 +175,10 @@ def emit_comparison_output(
     raise SystemExit(f"Unsupported emit mode: {emit}")
 
 
+def comparison_topic(entity_reports: list[tuple[str, schema.Report]]) -> str:
+    return " vs ".join(label for label, _ in entity_reports)
+
+
 def compute_save_path_display(save_dir: str, topic: str, suffix: str, emit: str) -> str:
     """Compute the user-friendly save path string that will be shown in the footer.
 
@@ -533,6 +541,13 @@ def main() -> int:
 
     config = env.get_config()
 
+    # Surface SSH-routing config as an env var so library modules (e.g.
+    # youtube_yt) can read it without taking a config dependency. This
+    # routes yt-dlp through `ssh <host>` to bypass YouTube's bot-wall on
+    # datacenter IPs (see lib/youtube_yt.py for details).
+    if config.get("LAST30DAYS_YOUTUBE_SSH_HOST") and "LAST30DAYS_YOUTUBE_SSH_HOST" not in os.environ:
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = config["LAST30DAYS_YOUTUBE_SSH_HOST"]
+
     # Handle setup subcommand
     topic = " ".join(args.topic).strip()
     if topic.lower() == "setup":
@@ -873,10 +888,15 @@ def main() -> int:
         pass
 
     fun_level = config.get("FUN_LEVEL", "medium").lower()
+    # Comparison HTML is the one case where the saved file's title and content
+    # have to be overridden away from the leading entity's report. Compute the
+    # gate once so the footer-display and save-output paths can't disagree.
+    is_comparison_html = bool(entity_reports) and args.emit == "html"
     footer_save_path = None
     if args.save_dir:
+        save_topic_for_display = comparison_topic(entity_reports) if is_comparison_html else report.topic
         footer_save_path = compute_save_path_display(
-            args.save_dir, report.topic, args.save_suffix or "", args.emit
+            args.save_dir, save_topic_for_display, args.save_suffix or "", args.emit
         )
 
     # Signal to render_compact whether pre-research flags were supplied.
@@ -917,6 +937,8 @@ def main() -> int:
             args.save_dir,
             suffix=args.save_suffix or "",
             synthesis_md=synthesis_md,
+            topic_override=comparison_topic(entity_reports) if is_comparison_html else None,
+            rendered_content=rendered if is_comparison_html else None,
         )
         sys.stderr.write(f"[last30days] Saved output to {save_path}\n")
         # Competitor / vs-mode: also save a per-entity raw file for each peer.

skills/last30days/scripts/lib/bird_x.py

@@ -9,6 +9,7 @@ import json
 import os
 import shutil
 import sys
+import time
 from pathlib import Path
 
 from . import http, log, subproc
@@ -17,6 +18,11 @@ from typing import Any, Dict, List, Optional, Tuple
 
 from .relevance import token_overlap_relevance as _compute_relevance
 
+# How many times to retry the bird-search subprocess when stdout is non-JSON
+# (typically an HTML anti-bot interstitial from Twitter's edge).
+MAX_JSON_DECODE_RETRIES = 2
+JSON_DECODE_RETRY_DELAY = 5.0  # seconds between retry attempts
+
 
 def _first_of(*values):
     """Return first value that is not None."""
@@ -148,16 +154,14 @@ def get_bird_status() -> Dict[str, Any]:
     }
 
 
-def _run_bird_search(query: str, count: int, timeout: int) -> Dict[str, Any]:
+def _invoke_bird_subprocess(query: str, count: int, timeout: int):
-    """Run a search using the vendored bird-search.mjs module.
+    """Invoke the vendored bird-search.mjs subprocess once.
 
-    Args:
+    Returns (result, error_dict). If error_dict is non-None, treat it as the
-        query: Full search query string (including since: filter)
+    final result and do not retry — those errors are terminal (timeout,
-        count: Number of results to request
+    spawn failure). If error_dict is None, the subprocess ran to completion
-        timeout: Timeout in seconds
+    and `result` is the SubprocResult; the caller decides whether to retry
-
+    based on the result.stdout content.
-    Returns:
-        Raw Bird JSON response or error dict.
     """
     cmd = [
         "node", str(_BIRD_SEARCH_MJS),
@@ -184,9 +188,9 @@ def _run_bird_search(query: str, count: int, timeout: int) -> Dict[str, Any]:
             on_pid=_register,
         )
     except subproc.SubprocTimeout:
-        return {"error": f"Search timed out after {timeout}s", "items": []}
+        return None, {"error": f"Search timed out after {timeout}s", "items": []}
     except Exception as e:
-        return {"error": str(e), "items": []}
+        return None, {"error": str(e), "items": []}
     finally:
         if pid_holder:
             try:
@@ -195,22 +199,80 @@ def _run_bird_search(query: str, count: int, timeout: int) -> Dict[str, Any]:
             except Exception:
                 pass
 
-    if result.returncode != 0:
+    return result, None
-        error = result.stderr.strip() or "Bird search failed"
-        return {"error": error, "items": []}
 
-    output = result.stdout.strip()
-    if not output:
-        return {"items": []}
 
-    try:
+def _run_bird_search(query: str, count: int, timeout: int) -> Dict[str, Any]:
-        parsed = json.loads(output)
+    """Run a search using the vendored bird-search.mjs module.
-    except json.JSONDecodeError as e:
-        return {"error": f"Invalid JSON response: {e}", "items": []}
 
-    if isinstance(parsed, list):
+    Retries the subprocess on JSON-decode failure (typically a Twitter
-        return {"items": parsed}
+    anti-bot HTML interstitial in stdout) up to MAX_JSON_DECODE_RETRIES
-    return parsed
+    times with JSON_DECODE_RETRY_DELAY seconds between attempts. Terminal
+    errors (subprocess timeout, non-zero return code) are returned
+    immediately without retry.
+
+    Args:
+        query: Full search query string (including since: filter)
+        count: Number of results to request
+        timeout: Timeout in seconds (per attempt)
+
+    Returns:
+        Raw Bird JSON response or error dict.
+    """
+    last_decode_error: Optional[str] = None
+
+    for attempt in range(MAX_JSON_DECODE_RETRIES):
+        result, terminal_error = _invoke_bird_subprocess(query, count, timeout)
+        if terminal_error is not None:
+            return terminal_error
+
+        if result.returncode != 0:
+            error = result.stderr.strip() or "Bird search failed"
+            return {"error": error, "items": []}
+
+        output = result.stdout.strip()
+        if not output:
+            return {"items": []}
+
+        try:
+            parsed = json.loads(output)
+        except json.JSONDecodeError as e:
+            # Twitter's edge sometimes serves an HTML anti-bot interstitial
+            # in place of JSON. Tag the failure shape so it's distinguishable
+            # from "no results" in logs, then retry the subprocess.
+            looks_html = output.lstrip().lower().startswith(("<!doctype", "<html", "<"))
+            attempt_num = attempt + 1
+            log_msg = (
+                f"Bird search returned non-JSON stdout "
+                f"(looks_html={looks_html}, attempt {attempt_num}/{MAX_JSON_DECODE_RETRIES}, "
+                f"first 80 chars: {output[:80]!r})"
+            )
+            last_decode_error = str(e)
+            if attempt_num < MAX_JSON_DECODE_RETRIES:
+                log.source_log(
+                    "X/bird",
+                    f"{log_msg}; retrying in {JSON_DECODE_RETRY_DELAY:.0f}s",
+                )
+                time.sleep(JSON_DECODE_RETRY_DELAY)
+                continue
+            log.source_log("X/bird", log_msg)
+            return {
+                "error": (
+                    f"Invalid JSON response after {MAX_JSON_DECODE_RETRIES} attempts "
+                    f"(likely Twitter anti-bot interstitial): {e}"
+                ),
+                "items": [],
+            }
+
+        if isinstance(parsed, list):
+            return {"items": parsed}
+        return parsed
+
+    # Defensive fallthrough — loop should always return above.
+    return {
+        "error": f"Bird search exhausted retries: {last_decode_error}",
+        "items": [],
+    }
 
 
 def search_x(

skills/last30days/scripts/lib/env.py

@@ -29,6 +29,23 @@ else:
 
 CODEX_AUTH_FILE = Path(os.environ.get("CODEX_AUTH_FILE", str(Path.home() / ".codex" / "auth.json")))
 
+# macOS Keychain integration: items stored with this service prefix are picked
+# up automatically on Darwin as the lowest-priority credential source.
+# Example: `security add-generic-password -a "$USER" -s last30days-XAI_API_KEY -w "xai-..."`.
+KEYCHAIN_SERVICE_PREFIX = "last30days-"
+
+# Single source of truth for which credentials the Keychain loader looks up.
+# The setup-keychain.sh helper mirrors this list and is held in sync via
+# tests/test_env_keychain.py::test_keychain_keys_match_setup_script.
+KEYCHAIN_KEYS = (
+    "OPENAI_API_KEY", "XAI_API_KEY", "GOOGLE_API_KEY", "GEMINI_API_KEY",
+    "GOOGLE_GENAI_API_KEY", "SCRAPECREATORS_API_KEY", "APIFY_API_TOKEN",
+    "AUTH_TOKEN", "CT0", "BSKY_HANDLE", "BSKY_APP_PASSWORD",
+    "TRUTHSOCIAL_TOKEN", "BRAVE_API_KEY", "EXA_API_KEY", "SERPER_API_KEY",
+    "OPENROUTER_API_KEY", "PARALLEL_API_KEY", "XQUIK_API_KEY",
+    "XIAOHONGSHU_API_BASE",
+)
+
 AuthSource = Literal["api_key", "codex", "none"]
 AuthStatus = Literal["ok", "missing", "expired", "missing_account_id"]
 
@@ -53,6 +70,10 @@ class OpenAIAuth:
 
 def _check_file_permissions(path: Path) -> None:
     """Warn to stderr if a secrets file has overly permissive permissions."""
+    if os.name == "nt":
+        # Windows reports synthesized POSIX mode bits that do not reflect NTFS ACLs.
+        return
+
     try:
         mode = path.stat().st_mode
         # Check if group or other can read (bits 0o044)
@@ -91,6 +112,46 @@ def load_env_file(path: Path) -> dict[str, str]:
     return env
 
 
+def _load_keychain(keys: list[str]) -> dict[str, str]:
+    """Load credentials from macOS Keychain (no-op on other platforms).
+
+    Each key is looked up as a generic password with service name
+    ``f"{KEYCHAIN_SERVICE_PREFIX}{key}"`` for the current user. Missing items
+    and lookup failures are silent — Keychain is the lowest-priority source
+    and is meant to be additive over `.env` files and process environment.
+    """
+    import platform
+    if platform.system() != "Darwin":
+        return {}
+
+    import shutil
+    security = shutil.which("security")
+    if not security:
+        return {}
+
+    import subprocess
+    import pwd
+    # USER can be unset under sudo, in Docker without --env USER, or in some CI
+    # runners; fall back to the OS user record so lookups still match items
+    # stored by setup-keychain.sh (which uses $USER).
+    user = os.environ.get("USER") or pwd.getpwuid(os.getuid()).pw_name
+    env: dict[str, str] = {}
+    for key in keys:
+        try:
+            result = subprocess.run(
+                [security, "find-generic-password",
+                 "-a", user,
+                 "-s", f"{KEYCHAIN_SERVICE_PREFIX}{key}",
+                 "-w"],
+                capture_output=True, text=True, timeout=5,
+            )
+        except (subprocess.TimeoutExpired, OSError):
+            continue
+        if result.returncode == 0 and result.stdout.strip():
+            env[key] = result.stdout.strip()
+    return env
+
+
 def _decode_jwt_payload(token: str) -> dict[str, Any] | None:
     """Decode JWT payload without verification."""
     try:
@@ -214,6 +275,7 @@ def get_config() -> dict[str, Any]:
       1. Environment variables (os.environ)
       2. .claude/last30days.env (per-project config)
       3. ~/.config/last30days/.env (global config)
+      4. macOS Keychain items prefixed ``last30days-`` (Darwin only)
     """
     # Load from global config file
     file_env = load_env_file(CONFIG_FILE) if CONFIG_FILE else {}
@@ -222,9 +284,14 @@ def get_config() -> dict[str, Any]:
     project_env_path = _find_project_env()
     project_env = load_env_file(project_env_path) if project_env_path else {}
 
-    # Merge: project overrides global
+    # Merge file sources: project > global
     merged_env = {**file_env, **project_env}
 
+    # Keychain is the lowest-priority source (Darwin only; no-op elsewhere).
+    # Loaded before openai_auth so OPENAI_API_KEY can come from Keychain too.
+    keychain_env = _load_keychain(list(KEYCHAIN_KEYS))
+    merged_env = {**keychain_env, **merged_env}
+
     openai_auth = get_openai_auth(merged_env)
 
     # Build config: Codex/OpenAI auth + process.env > project .env > global .env
@@ -265,16 +332,21 @@ def get_config() -> dict[str, Any]:
         ('FROM_BROWSER', None),
         ('SETUP_COMPLETE', None),
         ('INCLUDE_SOURCES', ''),
+        ('EXCLUDE_SOURCES', ''),
+        ('LAST30DAYS_YOUTUBE_SSH_HOST', None),
     ]
 
     for key, default in keys:
         config[key] = os.environ.get(key) or merged_env.get(key, default)
 
-    # Track which config source was used
+    # Track which config source was used (highest-priority file source wins
+    # the label; keychain is only reported when nothing else is configured).
     if project_env_path:
         config['_CONFIG_SOURCE'] = f'project:{project_env_path}'
     elif CONFIG_FILE and CONFIG_FILE.exists():
         config['_CONFIG_SOURCE'] = f'global:{CONFIG_FILE}'
+    elif keychain_env:
+        config['_CONFIG_SOURCE'] = 'keychain'
     else:
         config['_CONFIG_SOURCE'] = 'env_only'
 
@@ -517,12 +589,12 @@ def _parse_include_sources(config: dict[str, Any]) -> set[str]:
 def is_threads_available(config: dict[str, Any]) -> bool:
     """Check if Threads source is available.
 
-    Requires SCRAPECREATORS_API_KEY AND 'threads' in INCLUDE_SOURCES.
+    Returns True when SCRAPECREATORS_API_KEY is set. Threads runs alongside
-    Threads is an opt-in source - it is not activated by default.
+    TikTok and Instagram as part of the SC family — same key, same per-call
+    cost shape, so the same default-on rule applies. Suppress via
+    EXCLUDE_SOURCES=threads.
     """
-    if not config.get('SCRAPECREATORS_API_KEY'):
+    return bool(config.get('SCRAPECREATORS_API_KEY'))
-        return False
-    return 'threads' in _parse_include_sources(config)
 
 
 def is_instagram_available(config: dict[str, Any]) -> bool:

skills/last30days/scripts/lib/grounding.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import sys
 import urllib.parse
 from datetime import datetime
 from urllib.parse import urlparse
@@ -205,29 +206,90 @@ def web_search(
             backend = "parallel"
         else:
             return [], {}
+    items: list[dict] = []
+    artifact: dict = {}
     if backend == "brave":
         key = config.get("BRAVE_API_KEY")
         if not key:
             raise RuntimeError("BRAVE_API_KEY is required when web_backend='brave'")
-        return brave_search(query, date_range, key)
+        items, artifact = brave_search(query, date_range, key)
-    if backend == "exa":
+    elif backend == "exa":
         key = config.get("EXA_API_KEY")
         if not key:
             raise RuntimeError("EXA_API_KEY is required when web_backend='exa'")
-        return exa_search(query, date_range, key)
+        items, artifact = exa_search(query, date_range, key)
-    if backend == "serper":
+    elif backend == "serper":
         key = config.get("SERPER_API_KEY")
         if not key:
             raise RuntimeError("SERPER_API_KEY is required when web_backend='serper'")
-        return serper_search(query, date_range, key)
+        items, artifact = serper_search(query, date_range, key)
-    if backend == "parallel":
+    elif backend == "parallel":
         key = config.get("PARALLEL_API_KEY")
         if not key:
             raise RuntimeError("PARALLEL_API_KEY is required when web_backend='parallel'")
-        return parallel_search(query, date_range, key)
+        items, artifact = parallel_search(query, date_range, key)
-    if backend != "none":
+    elif backend != "none":
         raise ValueError(f"Unsupported web backend: {backend!r}")
-    return [], {}
+    else:
+        return [], {}
+    if items and not _reddit_excluded(config):
+        items = _enrich_reddit_items(items)
+    return items, artifact
+
+
+def _reddit_excluded(config: dict) -> bool:
+    """Return True when EXCLUDE_SOURCES contains 'reddit'.
+
+    Respects the same suppression knob the pipeline uses for source gating,
+    so a user who set EXCLUDE_SOURCES=reddit doesn't get Reddit content
+    smuggled back in via web-search URLs.
+    """
+    raw = (config.get("EXCLUDE_SOURCES") or "").split(",")
+    return any(s.strip().lower() == "reddit" for s in raw)
+
+
+def _enrich_reddit_items(items: list[dict]) -> list[dict]:
+    """Enrich web search results that are Reddit URLs with thread body and comments.
+
+    Claude Code's WebFetch blocks reddit.com, so the model can't retrieve
+    Reddit content from web search results. This fetches it via the public
+    JSON API (reddit.com/.../.json) which bypasses that restriction.
+
+    Callers should gate this with EXCLUDE_SOURCES=reddit handling (see
+    `_reddit_excluded`) so a user who explicitly excluded Reddit doesn't
+    get Reddit content via web-search URLs.
+    """
+    from . import reddit_enrich
+    from .reddit_enrich import RedditRateLimitError
+
+    for item in items:
+        url = item.get("url", "")
+        if "reddit.com" not in url or "/comments/" not in url:
+            continue
+        try:
+            thread_data = reddit_enrich.fetch_thread_data(url, timeout=8)
+            if not thread_data:
+                continue
+            parsed = reddit_enrich.parse_thread_data(thread_data)
+            # selftext lives under parsed["submission"], not at the top level
+            selftext = (parsed.get("submission") or {}).get("selftext", "")
+            if selftext:
+                item["snippet"] = selftext[:2000]
+            comments = parsed.get("comments", [])
+            top = reddit_enrich.get_top_comments(comments)
+            if top:
+                item["top_comments"] = [
+                    {"score": c.get("score", 0), "excerpt": (c.get("body") or "")[:200]}
+                    for c in top[:5]
+                ]
+            item["enriched_via"] = "reddit_json_api"
+        except RedditRateLimitError as exc:
+            # Stop iterating to avoid flooding more 429s
+            sys.stderr.write(f"[Web] Reddit rate-limited, halting enrichment: {exc}\n")
+            break
+        except Exception as exc:
+            sys.stderr.write(f"[Web] Reddit enrichment failed for {url}: {exc}\n")
+    return items
 
 
 # ---------------------------------------------------------------------------

skills/last30days/scripts/lib/hackernews.py

@@ -88,17 +88,26 @@ def search_hackernews(
 
     # Use extracted core subject instead of raw topic for cleaner Algolia matching
     core = extract_core_subject(topic)
-    _log(f"Searching for '{core}' (raw: '{topic}', since {from_date}, count={count})")
+    # Hyphens and commas tokenize awkwardly in Algolia; flatten them so themed
+    # queries like "ts-bun-node" or "claude, personal agents" become plain words.
+    core_flat = _flatten_query_for_algolia(core)
+    _log(f"Searching for '{core_flat}' (raw: '{topic}', since {from_date}, count={count})")
 
     # Use relevance-sorted search with minimum engagement filter.
     # NOTE: restrictSearchableAttributes=title omitted intentionally — it would
     # miss Ask HN/Show HN threads where the topic appears in the body.
     params = {
-        "query": core,
+        "query": core_flat,
         "tags": "story",
         "numericFilters": f"created_at_i>{from_ts},created_at_i<{to_ts},points>2",
         "hitsPerPage": str(count),
     }
+    # Algolia defaults to AND across query tokens, so a 4-5 word theme query
+    # matches no stories. Mark all-but-the-first token as optional so Algolia
+    # ranks by how many tokens match instead of requiring every one.
+    tokens = core_flat.split()
+    if len(tokens) > 1:
+        params["optionalWords"] = " ".join(tokens[1:])
 
     from urllib.parse import urlencode
     url = f"{ALGOLIA_SEARCH_URL}?{urlencode(params)}"
@@ -117,28 +126,56 @@ def search_hackernews(
     return response
 
 
-def _title_matches_query(title: str, query: str, author: str = "") -> bool:
+_WORD_BOUNDARY_RE_CACHE: Dict[str, "re.Pattern[str]"] = {}
-    """Check if the query term appears in the title content, not just an HN prefix or author.
 
-    Returns True if the query (or any multi-word token) appears in the title
+
-    after stripping "Tell HN:", "Show HN:", "Ask HN:", "Launch HN:" prefixes
+def _flatten_query_for_algolia(text: str) -> str:
-    and ignoring the author name.  Returns True when query is empty (no filter).
+    """Normalise query for Algolia + post-filter comparison.
+
+    Multi-keyword theme queries frequently contain commas (delimiters) or
+    hyphens (compound terms like ``ts-bun-node``); both tokenize awkwardly.
+    Flatten them to spaces and collapse runs of whitespace so the search
+    parameter and the post-filter operate on the same shape.
+    """
+    return " ".join(text.replace(",", " ").replace("-", " ").split())
+
+
+def _title_matches_query(title: str, query: str, author: str = "") -> bool:
+    """Check if any query token appears as a whole word in the title.
+
+    Returns True when the query is empty (no filter), or when at least one
+    query token matches as a whole word in the title after stripping
+    "Tell HN:", "Show HN:", "Ask HN:", "Launch HN:" prefixes.
+
+    We previously required *every* token to appear (all-words), which killed
+    every Algolia hit on multi-keyword themes like "claude, personal agents,
+    agentic infra" because real HN titles never contain all five tokens
+    verbatim. Relaxing to any-word matches Algolia's `optionalWords` behaviour
+    in `search_hackernews`. Token-overlap relevance scoring at parse time
+    demotes hits where only one weak token matched, so the loosened gate
+    won't surface noise to the top of the ranking.
+
+    Word-boundary matching (rather than naive substring) prevents short
+    tokens like ``ai`` or ``ts`` from matching unrelated words like
+    ``email`` or ``artists``.
     """
     if not query:
         return True
     stripped = _HN_PREFIXES.sub("", title).strip()
-    # Also check that the match isn't solely in the author's username
     check_text = stripped.lower()
-    query_lower = query.lower()
+    # Normalise the query the same way search_hackernews does so post-filter
-    # Check each word of the query independently; all must appear somewhere
+    # tokens line up with what Algolia actually saw.
-    # in the stripped title (not just the prefix).
+    query_words = [w for w in _flatten_query_for_algolia(query.lower()).split() if w]
-    query_words = query_lower.split()
+    if not query_words:
+        return True
     for word in query_words:
-        if word in check_text:
+        pattern = _WORD_BOUNDARY_RE_CACHE.get(word)
-            continue
+        if pattern is None:
-        # Word not found in stripped title — reject
+            pattern = re.compile(rf"\b{re.escape(word)}\b")
-        return False
+            _WORD_BOUNDARY_RE_CACHE[word] = pattern
-    return True
+        if pattern.search(check_text):
+            return True
+    return False
 
 
 def parse_hackernews_response(response: Dict[str, Any], query: str = "") -> List[Dict[str, Any]]:

skills/last30days/scripts/lib/http.py

@@ -2,6 +2,7 @@
 
 import json
 import re
+import socket
 import sys
 import time
 import urllib.error
@@ -22,9 +23,19 @@ def log(msg: str):
 MAX_RETRIES = 5
 MAX_429_RETRIES = 2
 RETRY_DELAY = 2.0
+# DNS resolution failures (gaierror) are transient — typically resolved by a
+# brief backoff and retry. Use a dedicated minimum attempt count + exponential
+# delays (1s, 2s, 4s) so callers that pass a small `retries` value still get a
+# meaningful chance to recover from a transient resolution failure.
+MIN_DNS_RETRIES = 3
 USER_AGENT = "last30days-skill/3.0 (Assistant Skill)"
 
 
+def _is_dns_failure(err: urllib.error.URLError) -> bool:
+    """Return True if a URLError was caused by DNS resolution (gaierror)."""
+    return isinstance(getattr(err, "reason", None), socket.gaierror)
+
+
 class HTTPError(Exception):
     """HTTP request error with status code."""
     def __init__(self, message: str, status_code: Optional[int] = None, body: Optional[str] = None):
@@ -85,7 +96,13 @@ def request(
 
     last_error = None
     rate_limit_count = 0
-    for attempt in range(retries):
+    # DNS failures get a dedicated minimum attempt count + exponential backoff.
+    # `effective_retries` is the actual loop bound; we expand it on the first
+    # gaierror if the caller passed a smaller `retries` value than MIN_DNS_RETRIES.
+    effective_retries = retries
+    dns_attempts = 0
+    attempt = 0
+    while attempt < effective_retries:
         try:
             with urllib.request.urlopen(req, timeout=timeout) as response:
                 body = response.read().decode('utf-8')
@@ -115,6 +132,8 @@ def request(
                 if rate_limit_count >= max_429_retries:
                     raise last_error
 
+            # HTTP errors respect the caller's original `retries`; only DNS
+            # failures get the widened `effective_retries` budget.
             if attempt < retries - 1:
                 if e.code == 429:
                     # Respect Retry-After header, fall back to exponential backoff
@@ -130,11 +149,43 @@ def request(
                 else:
                     delay = RETRY_DELAY * (2 ** attempt)
                 time.sleep(delay)
+            else:
+                # Caller's original retry budget exhausted; an earlier DNS
+                # failure may have widened `effective_retries`, but that
+                # widening is DNS-only — don't grant extra HTTP attempts.
+                break
         except urllib.error.URLError as e:
             log(f"URL Error: {e.reason}")
             last_error = HTTPError(f"URL Error: {e.reason}")
-            if attempt < retries - 1:
+            if _is_dns_failure(e):
+                # DNS resolution failures are transient; expand the retry budget
+                # to MIN_DNS_RETRIES if the caller passed fewer, and use
+                # exponential backoff (1s, 2s, 4s, ...) instead of the linear
+                # default. Counts DNS attempts separately so other URLError
+                # causes don't bypass the regular retry budget.
+                dns_attempts += 1
+                if effective_retries < MIN_DNS_RETRIES:
+                    log(
+                        f"DNS resolution failed; expanding retry budget from "
+                        f"{effective_retries} to {MIN_DNS_RETRIES}"
+                    )
+                    effective_retries = MIN_DNS_RETRIES
+                if attempt < effective_retries - 1:
+                    delay = 2 ** (dns_attempts - 1)  # 1s, 2s, 4s, 8s, ...
+                    log(
+                        f"DNS resolution failure (attempt {dns_attempts}); "
+                        f"retrying in {delay:.1f}s"
+                    )
+                    time.sleep(delay)
+            elif attempt < retries - 1:
+                # Non-DNS URLError (e.g. ConnectionRefused) respects the
+                # caller's original retry budget, not the DNS-widened bound.
                 time.sleep(RETRY_DELAY * (attempt + 1))
+            else:
+                # Caller's original retry budget exhausted; an earlier DNS
+                # failure widening `effective_retries` does not carry over
+                # to non-DNS error paths.
+                break
         except json.JSONDecodeError as e:
             log(f"JSON decode error: {e}")
             last_error = HTTPError(f"Invalid JSON response: {e}")
@@ -144,7 +195,13 @@ def request(
             log(f"Connection error: {type(e).__name__}: {e}")
             last_error = HTTPError(f"Connection error: {type(e).__name__}: {e}")
             if attempt < retries - 1:
+                # Socket errors respect the caller's original retry budget.
                 time.sleep(RETRY_DELAY * (attempt + 1))
+            else:
+                # Original budget exhausted; DNS widening doesn't apply here.
+                break
+
+        attempt += 1
 
     if last_error:
         raise last_error

skills/last30days/scripts/lib/pipeline.py

@@ -128,6 +128,9 @@ def available_sources(config: dict[str, Any], requested_sources: list[str] | Non
         available.append("pinterest")
     if env.is_xquik_available(config):
         available.append("xquik")
+    exclude = {s.strip().lower() for s in (config.get("EXCLUDE_SOURCES") or "").split(",") if s.strip()}
+    if exclude:
+        available = [s for s in available if s not in exclude]
     return available
 
 

skills/last30days/scripts/lib/providers.py

@@ -9,7 +9,7 @@ from typing import Any
 
 from . import env, http, schema
 
-GEMINI_FLASH_LITE = "gemini-3.1-flash-lite-preview"
+GEMINI_FLASH_LITE = "gemini-3.1-flash-lite"
 GEMINI_PRO = "gemini-3.1-pro-preview"
 OPENAI_DEFAULT = "gpt-5.4-nano"
 XAI_DEFAULT = "grok-4-1-fast"
@@ -232,8 +232,8 @@ def _resolve_model_pins(config: dict[str, Any], depth: str, provider_name: str)
     rerank_model = config.get("LAST30DAYS_RERANK_MODEL") or default_rerank
 
     if provider_name == "gemini":
-        _require_gemini_31_preview(planner_model, role="planner")
+        _require_gemini_31(planner_model, role="planner")
-        _require_gemini_31_preview(rerank_model, role="rerank")
+        _require_gemini_31(rerank_model, role="rerank")
 
     return planner_model, rerank_model
 
@@ -344,11 +344,11 @@ def _resolve_x_backend(config: dict[str, Any]) -> str | None:
     return env.get_x_source(config)
 
 
-def _require_gemini_31_preview(model: str, *, role: str) -> None:
+def _require_gemini_31(model: str, *, role: str) -> None:
-    if model.startswith("gemini-3.1-") and model.endswith("-preview"):
+    if model.startswith("gemini-3.1-"):
         return
     raise RuntimeError(
-        f"{role} must use a Gemini 3.1 preview model. Got: {model}"
+        f"{role} must use a Gemini 3.1 model. Got: {model}"
     )
 
 

skills/last30days/scripts/lib/render.py

@@ -4,18 +4,11 @@ from __future__ import annotations
 
 import json
 import pathlib
-import re
 from collections import Counter
 from datetime import date
 from urllib.parse import urlparse
 
-from . import dates, schema
+from . import dates, schema, skill_meta
-
-
-_VERSION_RE = re.compile(
-    r'''^version:\s*(?:"([^"]+)"|'([^']+)'|(\S+))\s*$''',
-    re.MULTILINE,
-)
 
 
 def _skill_version() -> str:
@@ -25,11 +18,12 @@ def _skill_version() -> str:
     Hermes, etc.) do not always carry `.claude-plugin/plugin.json` — that file ships with
     plugin-cache installs but not with per-harness skill installs. SKILL.md frontmatter is
     the fallback that keeps the badge from emitting v? on those installs. Returns "?" only
-    if both sources are missing.
+    if no usable version string is found from either source (missing files, corrupt JSON,
+    or SKILL.md without a version line).
 
     A corrupt manifest at one ancestor does not shadow a valid manifest at a deeper one
-    (continue, not break). YAML frontmatter accepts double-quoted, single-quoted, or
+    (continue, not break). SKILL.md parsing accepts double-quoted, single-quoted, or
-    unquoted version scalars.
+    unquoted YAML version scalars (delegated to skill_meta.read_skill_version).
     """
     here = pathlib.Path(__file__).resolve()
     for parent in here.parents:
@@ -43,16 +37,11 @@ def _skill_version() -> str:
                 return version
 
     # No usable manifest found at any ancestor — fall back to SKILL.md frontmatter.
+    # First SKILL.md found in the walk is THIS skill's; never traverse past it.
     for parent in here.parents:
         skill_md = parent / "SKILL.md"
         if skill_md.is_file():
-            try:
+            return skill_meta.read_skill_version(skill_md) or "?"
-                match = _VERSION_RE.search(skill_md.read_text())
-            except (OSError, UnicodeDecodeError):
-                break
-            if match:
-                return next(g for g in match.groups() if g is not None)
-            break
     return "?"
 
 
@@ -107,7 +96,7 @@ def render_compact(report: schema.Report, cluster_limit: int = 8, fun_level: str
     non_empty = [s for s, items in sorted(report.items_by_source.items()) if items]
     lines = [
         *_render_badge(),
-        f"# last30days v3.0.0: {report.topic}",
+        f"# last30days v{_skill_version()}: {report.topic}",
         "",
         *_assistant_safety_lines(),
         f"- Date range: {report.range_from} to {report.range_to}",
@@ -613,7 +602,7 @@ def render_comparison_multi(
 
     lines: list[str] = [
         *_render_badge(),
-        f"# last30days v3.0.0: {synthesized_topic}",
+        f"# last30days v{_skill_version()}: {synthesized_topic}",
         "",
         *_assistant_safety_lines(),
         f"- Comparison mode: {len(entities)} entities ({', '.join(entities)})",
@@ -801,7 +790,7 @@ def render_full(report: schema.Report) -> str:
     # Start with the same header as compact
     non_empty = [s for s, items in sorted(report.items_by_source.items()) if items]
     lines = [
-        f"# last30days v3.0.0: {report.topic}",
+        f"# last30days v{_skill_version()}: {report.topic}",
         "",
         *_assistant_safety_lines(),
         f"- Date range: {report.range_from} to {report.range_to}",

skills/last30days/scripts/lib/setup_wizard.py

@@ -335,8 +335,9 @@ def poll_device_auth(
     """
     import sys
 
-    deadline = time.time() + timeout
+    started_at = time.time()
-    last_reminder = time.time()
+    deadline = started_at + timeout
+    last_reminder = started_at
     reminder_count = 0
     max_reminders = 4
     reminder_interval = 30  # seconds between reminders

skills/last30days/scripts/lib/skill_meta.py

@@ -0,0 +1,33 @@
+"""SKILL.md metadata helpers — single source of truth for parsing skill frontmatter.
+
+Centralizes the version regex that previously lived in render.py and was
+duplicated in tests/test_plugin_contract.py and tests/test_version_consistency.py.
+"""
+
+import re
+from pathlib import Path
+
+# Matches `version: "x.y.z"`, `version: 'x.y.z'`, or `version: x.y.z` in YAML
+# frontmatter. Multiline so the pattern can be applied to a full SKILL.md text.
+# Three alternation groups — exactly one captures per successful match.
+_VERSION_RE = re.compile(
+    r'''^version:\s*(?:"([^"]+)"|'([^']+)'|(\S+))\s*$''',
+    re.MULTILINE,
+)
+
+
+def read_skill_version(skill_md_path: Path) -> str | None:
+    """Return the version string from a SKILL.md's frontmatter, or None.
+
+    Returns None if the file can't be read (missing, permission, decode error)
+    or if no `version:` line is found. Accepts double-quoted, single-quoted,
+    or unquoted YAML version scalars.
+    """
+    try:
+        text = skill_md_path.read_text()
+    except (OSError, UnicodeDecodeError):
+        return None
+    match = _VERSION_RE.search(text)
+    if not match:
+        return None
+    return match.group(1) or match.group(2) or match.group(3)

skills/last30days/scripts/lib/ui.py

@@ -6,6 +6,8 @@ import threading
 import random
 from typing import Optional
 
+from .render import _skill_version
+
 # Check if we're in a real terminal (not captured by Claude Code)
 IS_TTY = sys.stderr.isatty()
 
@@ -198,7 +200,7 @@ Just start with "last30" and talk to me like normal.
 
 # Shorter promo for single missing key
 PROMO_SINGLE_KEY = {
-    "reddit": "\n💡 Unlock TikTok and Instagram with SCRAPECREATORS_API_KEY - 10,000 free calls, no CC - scrapecreators.com\n",
+    "reddit": "\n💡 Unlock TikTok and Instagram with SCRAPECREATORS_API_KEY - 100 free credits, no CC - scrapecreators.com\n",
     "x": "\n💡 Unlock X: log into x.com in Firefox or Safari, then re-run. Or add AUTH_TOKEN/CT0 or XAI_API_KEY.\n",
     "web": "\n💡 You can unlock native grounded web search with BRAVE_API_KEY or SERPER_API_KEY.\n",
 }
@@ -509,7 +511,8 @@ def show_diagnostic_banner(diag: dict):
 
     if IS_TTY:
         lines.append(f"{Colors.DIM}┌─────────────────────────────────────────────────────┐{Colors.RESET}")
-        lines.append(f"{Colors.DIM}│{Colors.RESET} {Colors.BOLD}/last30days v3.0.0 - Source Status{Colors.RESET}                 {Colors.DIM}│{Colors.RESET}")
+        _header = f"/last30days v{_skill_version()} - Source Status"
+        lines.append(f"{Colors.DIM}│{Colors.RESET} {Colors.BOLD}{_header}{Colors.RESET}{' ' * (52 - len(_header))}{Colors.DIM}│{Colors.RESET}")
         lines.append(f"{Colors.DIM}│{Colors.RESET}                                                     {Colors.DIM}│{Colors.RESET}")
 
         # Reddit
@@ -556,7 +559,8 @@ def show_diagnostic_banner(diag: dict):
     else:
         # Plain text for non-TTY (Claude Code / Codex)
         lines.append("┌─────────────────────────────────────────────────────┐")
-        lines.append("│ /last30days v3.0.0 - Source Status                 │")
+        _header_plain = f"/last30days v{_skill_version()} - Source Status"
+        lines.append(f"│ {_header_plain}{' ' * (52 - len(_header_plain))}│")
         lines.append("│                                                     │")
 
         if has_reddit and has_scrapecreators:

skills/last30days/scripts/lib/youtube_yt.py

@@ -8,7 +8,9 @@ Inspired by Peter Steinberger's toolchain approach (yt-dlp + summarize CLI).
 
 import json
 import math
+import os
 import re
+import shlex
 import shutil
 import sys
 import tempfile
@@ -96,10 +98,76 @@ def _log(msg: str):
 
 
 def is_ytdlp_installed() -> bool:
-    """Check if yt-dlp is available in PATH."""
+    """Check if yt-dlp is available locally, or if SSH routing is configured.
+
+    When LAST30DAYS_YOUTUBE_SSH_HOST is set, returns True without a local check —
+    yt-dlp lives on the remote host. Failures surface naturally on first use.
+    """
+    if _ytdlp_ssh_host():
+        return True
     return shutil.which("yt-dlp") is not None
 
 
+# Host aliases must be plain hostnames / SSH config aliases — no flags, no
+# shell metacharacters. Rejects any value that could be reinterpreted by ssh
+# (or the surrounding shell) as something other than a destination.
+_SSH_HOST_ALIAS_RE = re.compile(r"^[a-zA-Z0-9._-]+$")
+
+
+def _ytdlp_ssh_host() -> Optional[str]:
+    """Return SSH host alias if yt-dlp should be routed via SSH, else None.
+
+    Set LAST30DAYS_YOUTUBE_SSH_HOST=<ssh-alias> (e.g. 'macmini') in the environment
+    to route yt-dlp through SSH for residential IP egress. This bypasses
+    YouTube's bot-wall on datacenter IPs (Hetzner, DigitalOcean, AWS, etc.)
+    where ytsearch returns 0 results regardless of cookies.
+
+    The remote host must have yt-dlp installed and reachable via the named
+    SSH alias (configured in ~/.ssh/config). On macOS hosts with Homebrew,
+    add brew shellenv to ~/.zshenv (not just ~/.zprofile) so non-login SSH
+    shells find yt-dlp on PATH.
+
+    Validation: host value must match ``[A-Za-z0-9._-]+``. Anything starting
+    with ``-`` or containing shell/SSH metacharacters is rejected with a
+    stderr warning and treated as unset, so a misconfigured or attacker-
+    controlled value can't slip through as an SSH option flag or proxy command.
+    The ``--`` option terminator in ``_wrap_ytdlp_cmd`` is a second line of
+    defense; this regex closes the door on the env var ever reaching ssh
+    in the first place.
+
+    To use a value from ~/.config/last30days/.env, export it into the
+    environment before invoking the engine, e.g. in a wrapper:
+        set -a; source ~/.config/last30days/.env; set +a
+        python3 last30days.py "..."
+    """
+    host = os.environ.get("LAST30DAYS_YOUTUBE_SSH_HOST", "").strip()
+    if not host:
+        return None
+    if not _SSH_HOST_ALIAS_RE.match(host):
+        sys.stderr.write(
+            f"[youtube_yt] WARNING: LAST30DAYS_YOUTUBE_SSH_HOST={host!r} "
+            "does not look like a plain hostname/alias; ignoring. "
+            "Expected pattern: letters, digits, dot, underscore, hyphen.\n"
+        )
+        return None
+    return host
+
+
+def _wrap_ytdlp_cmd(cmd: List[str]) -> List[str]:
+    """Wrap a yt-dlp command list with `ssh <host>` when SSH routing is set.
+
+    Args are shell-quoted to survive the remote shell. Uses BatchMode=yes so
+    a misconfigured key fails fast instead of hanging on a password prompt.
+    The `--` option terminator prevents an SSH option-injection if
+    LAST30DAYS_YOUTUBE_SSH_HOST were ever set to a value starting with `-`.
+    """
+    host = _ytdlp_ssh_host()
+    if not host:
+        return cmd
+    remote_cmd = " ".join(shlex.quote(a) for a in cmd)
+    return ["ssh", "-o", "BatchMode=yes", "--", host, remote_cmd]
+
+
 def _extract_core_subject(topic: str) -> str:
     """Extract core subject from verbose query for YouTube search.
 
@@ -223,6 +291,7 @@ def search_youtube(
         "--no-warnings",
         "--no-download",
     ]
+    cmd = _wrap_ytdlp_cmd(cmd)
 
     try:
         result = subproc.run_with_timeout(cmd, timeout=120)
@@ -472,13 +541,22 @@ def fetch_transcript(video_id: str, temp_dir: str) -> Optional[str]:
         Plaintext transcript string, or None if no captions available.
     """
     raw_vtt = None
-    if is_ytdlp_installed():
+    # When SSH-routing is on, the yt-dlp transcript path would write a VTT
+    # file on the remote host that we can't easily read back. Skip it and
+    # use the HTTP transcript fallback (different YouTube endpoint, less
+    # bot-walled, works fine from datacenter IPs).
+    ssh_host = _ytdlp_ssh_host()
+    use_ytdlp = is_ytdlp_installed() and not ssh_host
+    if use_ytdlp:
         raw_vtt = _fetch_transcript_ytdlp(video_id, temp_dir)
         if not raw_vtt:
             _log(f"yt-dlp transcript failed for {video_id}, trying direct HTTP fallback")
             raw_vtt = _fetch_transcript_direct(video_id)
     else:
-        _log("yt-dlp not installed, using direct HTTP transcript fetch")
+        if ssh_host:
+            _log("SSH-routing active, using direct HTTP transcript fetch")
+        else:
+            _log("yt-dlp not installed, using direct HTTP transcript fetch")
         raw_vtt = _fetch_transcript_direct(video_id)
 
     if not raw_vtt:
@@ -866,9 +944,12 @@ def _sc_youtube_search(keyword: str, token: str) -> List[Dict[str, Any]]:
         List of raw video dicts from the API.
     """
     try:
+        # SC's /v1/youtube/search rejects ?keyword= with HTTP 400; the canonical
+        # parameter for that endpoint is `query`. Other SC endpoints use their
+        # own per-endpoint param names so this was the lone outlier.
         data = http.get(
             f"{SCRAPECREATORS_YT_BASE}/search",
-            params={"keyword": keyword},
+            params={"query": keyword},
             headers=http.scrapecreators_headers(token),
             timeout=30,
             retries=2,

skills/last30days/scripts/setup-keychain.sh

@@ -0,0 +1,122 @@
+#!/bin/bash
+# Store last30days API keys in the macOS Keychain.
+#
+# Keys are stored as generic passwords with service name `last30days-<KEY>`
+# for the current user. The lib/env.py loader picks them up automatically as
+# the lowest-priority credential source on Darwin.
+#
+# Usage:
+#   ./setup-keychain.sh              # interactive: prompts for each key
+#   ./setup-keychain.sh KEY [KEY..]  # prompt only for the listed keys
+#   ./setup-keychain.sh --list       # list which last30days-* items exist
+#   ./setup-keychain.sh --delete KEY # remove a stored key
+#
+# Existing values are shown as "(set)" and skipped unless --replace is passed.
+# Skip any prompt with empty input.
+
+set -euo pipefail
+
+PREFIX="last30days-"
+# Mirrors lib/env.py::KEYCHAIN_KEYS — kept in sync via
+# tests/test_env_keychain.py::test_keychain_keys_match_setup_script.
+ALL_KEYS=(
+  OPENAI_API_KEY
+  XAI_API_KEY
+  GOOGLE_API_KEY
+  GEMINI_API_KEY
+  GOOGLE_GENAI_API_KEY
+  SCRAPECREATORS_API_KEY
+  APIFY_API_TOKEN
+  AUTH_TOKEN
+  CT0
+  BSKY_HANDLE
+  BSKY_APP_PASSWORD
+  TRUTHSOCIAL_TOKEN
+  BRAVE_API_KEY
+  EXA_API_KEY
+  SERPER_API_KEY
+  OPENROUTER_API_KEY
+  PARALLEL_API_KEY
+  XQUIK_API_KEY
+  XIAOHONGSHU_API_BASE
+)
+
+if [[ "${OSTYPE:-}" != darwin* ]]; then
+  echo "setup-keychain.sh requires macOS (security command). Got: $OSTYPE" >&2
+  exit 1
+fi
+if ! command -v security >/dev/null 2>&1; then
+  echo "security command not found on PATH" >&2
+  exit 1
+fi
+
+REPLACE=0
+ACTION="prompt"
+TARGETS=()
+
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --list) ACTION="list"; shift ;;
+    --delete) ACTION="delete"; shift ;;
+    --replace) REPLACE=1; shift ;;
+    --help|-h) sed -n '2,/^$/p' "$0" | sed 's/^# //; s/^#//'; exit 0 ;;
+    -*) echo "unknown flag: $1" >&2; exit 2 ;;
+    *) TARGETS+=("$1"); shift ;;
+  esac
+done
+
+case "$ACTION" in
+  list)
+    echo "Stored last30days-* keychain items:"
+    for key in "${ALL_KEYS[@]}"; do
+      if security find-generic-password -a "$USER" -s "${PREFIX}${key}" -w >/dev/null 2>&1; then
+        echo "  $key"
+      fi
+    done
+    exit 0
+    ;;
+  delete)
+    if [[ ${#TARGETS[@]} -eq 0 ]]; then
+      echo "--delete needs at least one KEY name" >&2; exit 2
+    fi
+    for key in "${TARGETS[@]}"; do
+      if security delete-generic-password -a "$USER" -s "${PREFIX}${key}" >/dev/null 2>&1; then
+        echo "deleted: $key"
+      else
+        echo "not found: $key"
+      fi
+    done
+    exit 0
+    ;;
+esac
+
+if [[ ${#TARGETS[@]} -eq 0 ]]; then
+  TARGETS=("${ALL_KEYS[@]}")
+fi
+
+added=0; skipped=0; replaced=0
+for key in "${TARGETS[@]}"; do
+  existing="$(security find-generic-password -a "$USER" -s "${PREFIX}${key}" -w 2>/dev/null || true)"
+  if [[ -n "$existing" && "$REPLACE" -eq 0 ]]; then
+    printf "  %-28s (set, skipping — use --replace to overwrite)\n" "$key"
+    skipped=$((skipped + 1))
+    continue
+  fi
+  printf "  %-28s " "$key"
+  IFS= read -rs value
+  echo
+  if [[ -z "$value" ]]; then
+    skipped=$((skipped + 1))
+    continue
+  fi
+  security add-generic-password -U -a "$USER" -s "${PREFIX}${key}" -w "$value"
+  if [[ -n "$existing" ]]; then
+    replaced=$((replaced + 1))
+  else
+    added=$((added + 1))
+  fi
+done
+
+echo
+echo "Done. added=$added replaced=$replaced skipped=$skipped"
+echo "Verify with: $0 --list"

skills/last30days/scripts/store.py

@@ -14,7 +14,7 @@ import argparse
 import json
 import sqlite3
 import sys
-from datetime import datetime, timedelta
+from datetime import datetime, timedelta, timezone
 from pathlib import Path
 from typing import Any, Dict, List, Optional
 
@@ -159,7 +159,30 @@ _UPDATABLE_FINDING_COLUMNS = frozenset({
 })
 
 # Future migrations keyed by version number
-MIGRATIONS: Dict[int, str] = {}
+MIGRATIONS: Dict[int, str] = {
+    2: """
+CREATE TABLE IF NOT EXISTS finding_sightings (
+    id INTEGER PRIMARY KEY,
+    finding_id INTEGER NOT NULL REFERENCES findings(id) ON DELETE CASCADE,
+    run_id INTEGER REFERENCES research_runs(id) ON DELETE CASCADE,
+    topic_id INTEGER REFERENCES topics(id) ON DELETE CASCADE,
+    source TEXT NOT NULL,
+    source_url TEXT NOT NULL,
+    source_title TEXT,
+    engagement_score REAL,
+    relevance_score REAL,
+    seen_at TEXT DEFAULT (datetime('now')),
+    UNIQUE(run_id, finding_id)
+);
+
+CREATE INDEX IF NOT EXISTS idx_finding_sightings_run
+    ON finding_sightings(run_id, topic_id);
+CREATE INDEX IF NOT EXISTS idx_finding_sightings_topic_seen
+    ON finding_sightings(topic_id, seen_at);
+CREATE INDEX IF NOT EXISTS idx_finding_sightings_url
+    ON finding_sightings(source_url);
+""",
+}
 
 
 def _connect(db_path: Optional[Path] = None) -> sqlite3.Connection:
@@ -423,6 +446,7 @@ def store_findings(
 
         new_count = len(insert_rows)
         updated_count = len(update_rows)
+        _record_sightings(conn, run_id, topic_id, with_urls, existing_by_url)
         conn.execute(
             "UPDATE research_runs SET findings_new = ?, findings_updated = ? WHERE id = ?",
             (new_count, updated_count, run_id),
@@ -434,6 +458,84 @@ def store_findings(
     return {"new": new_count, "updated": updated_count}
 
 
+def _record_sightings(
+    conn: sqlite3.Connection,
+    run_id: int,
+    topic_id: int,
+    findings_with_urls: List[tuple[str, Dict[str, Any]]],
+    existing_by_url: Optional[Dict[str, sqlite3.Row]] = None,
+) -> None:
+    """Record the findings observed during this run.
+
+    The aggregate findings table keeps one row per URL and updates that row on
+    re-sighting. This ledger preserves the run/topic membership needed for
+    watchlist deltas and dossiers.
+    """
+    if not findings_with_urls:
+        return
+
+    by_url = {url: finding for url, finding in findings_with_urls}
+    rows_by_url = dict(existing_by_url or {})
+
+    missing_urls = [url for url in by_url if url not in rows_by_url]
+    if missing_urls:
+        placeholders = ",".join("?" for _ in missing_urls)
+        rows = conn.execute(
+            f"SELECT id, source_url FROM findings WHERE source_url IN ({placeholders})",
+            missing_urls,
+        ).fetchall()
+        rows_by_url.update({row["source_url"]: row for row in rows})
+
+    sighting_rows = []
+    for url, finding in by_url.items():
+        row = rows_by_url.get(url)
+        if row is None:
+            continue
+        sighting_rows.append((
+            row["id"],
+            run_id,
+            topic_id,
+            finding.get("source", "unknown"),
+            url,
+            finding.get("source_title") or finding.get("title", ""),
+            finding.get("engagement_score", 0),
+            finding.get("relevance_score", 0),
+        ))
+
+    if not sighting_rows:
+        return
+
+    conn.executemany(
+        """INSERT INTO finding_sightings
+           (finding_id, run_id, topic_id, source, source_url, source_title,
+            engagement_score, relevance_score)
+           VALUES (?, ?, ?, ?, ?, ?, ?, ?)
+           ON CONFLICT(run_id, finding_id) DO UPDATE SET
+             topic_id = excluded.topic_id,
+             source = excluded.source,
+             source_url = excluded.source_url,
+             source_title = excluded.source_title,
+             engagement_score = excluded.engagement_score,
+             relevance_score = excluded.relevance_score""",
+        sighting_rows,
+    )
+
+
+def get_sightings_for_run(topic_id: int, run_id: int) -> List[Dict[str, Any]]:
+    """Return findings observed for a topic during a specific run."""
+    conn = _connect()
+    try:
+        rows = conn.execute(
+            """SELECT * FROM finding_sightings
+               WHERE topic_id = ? AND run_id = ?
+               ORDER BY id""",
+            (topic_id, run_id),
+        ).fetchall()
+        return [dict(r) for r in rows]
+    finally:
+        conn.close()
+
+
 def get_new_findings(
     topic_id: int,
     since: Optional[str] = None,
@@ -519,7 +621,7 @@ def get_daily_cost(date: Optional[str] = None) -> float:
     conn = _connect()
     try:
         if not date:
-            date = datetime.now().strftime("%Y-%m-%d")
+            date = datetime.now(timezone.utc).strftime("%Y-%m-%d")
         row = conn.execute(
             """SELECT COALESCE(SUM(token_cost), 0) as total
                FROM research_runs
@@ -575,7 +677,7 @@ def get_stats() -> Dict[str, Any]:
         topic_count = conn.execute("SELECT COUNT(*) FROM topics WHERE enabled = 1").fetchone()[0]
         finding_count = conn.execute("SELECT COUNT(*) FROM findings").fetchone()[0]
 
-        week_ago = (datetime.now() - timedelta(days=7)).strftime("%Y-%m-%d")
+        week_ago = (datetime.now(timezone.utc) - timedelta(days=7)).strftime("%Y-%m-%d")
         runs_7d = conn.execute(
             "SELECT COUNT(*) FROM research_runs WHERE run_date >= ?", (week_ago,)
         ).fetchone()[0]
@@ -621,7 +723,7 @@ def get_trending(days: int = 7) -> List[Dict[str, Any]]:
     """Get topics ranked by recent finding activity."""
     conn = _connect()
     try:
-        since = (datetime.now() - timedelta(days=days)).strftime("%Y-%m-%d")
+        since = (datetime.now(timezone.utc) - timedelta(days=days)).strftime("%Y-%m-%d")
         rows = conn.execute(
             """SELECT t.name, t.id,
                       COUNT(f.id) as new_findings,
@@ -673,27 +775,31 @@ def findings_from_report(
     limit: Optional[int] = None,
 ) -> List[Dict[str, Any]]:
     """Convert report into persisted findings.
-    
+
     Uses ranked candidates (post-rerank) when available for quality scores and explanations.
     Supplements with raw items from items_by_source for HN/PM that didn't rank highly
-    but are valuable for watchlist persistence.
+    but are valuable for watchlist persistence. When ranked_candidates is empty
+    (degraded path — rerank failed or was skipped), falls back to supplementing
+    all sources from items_by_source so findings aren't silently dropped.
     """
     findings = []
     seen_urls = set()
-    
+
-    # Phase 1: Process ranked candidates (high-quality data with explanations and corroboration)
     for candidate in report.ranked_candidates:
-        finding = finding_from_candidate(candidate)
+        findings.append(finding_from_candidate(candidate))
-        findings.append(finding)
         seen_urls.add(candidate.url)
-    
+
-    # Phase 2: Add HN/PM items not already captured in ranked candidates
+    supplement_sources = (
-    for source_name in ["hackernews", "polymarket"]:
+        list(report.items_by_source)
+        if not report.ranked_candidates
+        else ["hackernews", "polymarket"]
+    )
+    for source_name in supplement_sources:
         if source_name not in report.items_by_source:
             continue
         for item in report.items_by_source[source_name]:
             if item.url in seen_urls:
-                continue  # Already captured with rich data
+                continue
             findings.append({
                 "source": source_name,
                 "source_url": item.url,
@@ -705,8 +811,7 @@ def findings_from_report(
                 "relevance_score": item.local_relevance or 0.5,
             })
             seen_urls.add(item.url)
-    
+
-    # Apply global limit after collecting all findings (fix: was per-source, now global)
     return findings[:limit] if limit is not None else findings
 
 
@@ -722,9 +827,10 @@ def _cli_query(args):
 
     since = None
     if args.since:
-        # Parse duration like "7d", "30d"
+        # Parse duration like "7d", "30d". Use UTC to match SQLite's
+        # datetime('now') which writes first_seen in UTC.
         days = int(args.since.rstrip("d"))
-        since = (datetime.now() - timedelta(days=days)).strftime("%Y-%m-%d")
+        since = (datetime.now(timezone.utc) - timedelta(days=days)).strftime("%Y-%m-%d")
 
     findings = get_new_findings(topic["id"], since)
     print(json.dumps({"topic": topic["name"], "findings": findings, "count": len(findings)}, default=str))

tests/test_bird_x.py

@@ -232,5 +232,79 @@ class TestVendoredBirdRuntime(unittest.TestCase):
         self.assertEqual(5, items[0]["engagement"]["likes"])
 
 
+class TestRunBirdSearchJsonDecodeRetry(unittest.TestCase):
+    """When bird-search returns non-JSON stdout, retry the subprocess.
+
+    Twitter's edge sometimes serves an HTML anti-bot interstitial in place of
+    JSON. Before this fix, that response made json.loads raise JSONDecodeError
+    and the function returned {"items": []} with no diagnostic — silent-empty
+    against an orchestrator that can't distinguish "Twitter blocked us" from
+    "no tweets matched the query."
+    """
+
+    def _make_result(self, stdout: str, stderr: str = "", returncode: int = 0):
+        from lib.subproc import SubprocResult
+        return SubprocResult(returncode=returncode, stdout=stdout, stderr=stderr)
+
+    def test_retries_subprocess_on_html_interstitial_then_succeeds(self):
+        """First subprocess attempt returns HTML; second returns JSON → success."""
+        from unittest import mock
+        from lib import bird_x
+
+        html_interstitial = "<!DOCTYPE html><html><body>Rate limited</body></html>"
+        json_success = '[{"id": "1", "text": "tweet"}]'
+
+        results = [
+            (self._make_result(stdout=html_interstitial), None),
+            (self._make_result(stdout=json_success), None),
+        ]
+
+        with mock.patch.object(bird_x, "_invoke_bird_subprocess", side_effect=results), \
+             mock.patch.object(bird_x.time, "sleep") as mock_sleep:
+            response = bird_x._run_bird_search("test", count=10, timeout=30)
+
+        self.assertNotIn("error", response)
+        self.assertEqual(response["items"], [{"id": "1", "text": "tweet"}])
+        # Should have slept between the failed first attempt and the retry.
+        mock_sleep.assert_called_once_with(bird_x.JSON_DECODE_RETRY_DELAY)
+
+    def test_returns_error_after_all_retries_exhausted(self):
+        """All attempts return HTML → error dict with diagnostic + items=[]."""
+        from unittest import mock
+        from lib import bird_x
+
+        html_interstitial = "<!DOCTYPE html><html>blocked</html>"
+        results = [
+            (self._make_result(stdout=html_interstitial), None),
+            (self._make_result(stdout=html_interstitial), None),
+        ]
+
+        with mock.patch.object(bird_x, "_invoke_bird_subprocess", side_effect=results), \
+             mock.patch.object(bird_x.time, "sleep"):
+            response = bird_x._run_bird_search("test", count=10, timeout=30)
+
+        self.assertIn("error", response)
+        self.assertIn("Invalid JSON response", response["error"])
+        # Diagnostic message names the anti-bot interstitial so it's
+        # distinguishable from a genuine no-results case in logs.
+        self.assertIn("anti-bot interstitial", response["error"].lower())
+        self.assertEqual(response["items"], [])
+
+    def test_terminal_subprocess_error_is_not_retried(self):
+        """Subprocess timeout / spawn failure → terminal error, no retry."""
+        from unittest import mock
+        from lib import bird_x
+
+        timeout_error = {"error": "Search timed out after 30s", "items": []}
+        results = [(None, timeout_error)]
+
+        with mock.patch.object(bird_x, "_invoke_bird_subprocess", side_effect=results), \
+             mock.patch.object(bird_x.time, "sleep") as mock_sleep:
+            response = bird_x._run_bird_search("test", count=10, timeout=30)
+
+        self.assertEqual(response, timeout_error)
+        mock_sleep.assert_not_called()
+
+
 if __name__ == "__main__":
     unittest.main()

tests/test_cli_v3.py

@@ -27,8 +27,8 @@ class CliV3Tests(unittest.TestCase):
             generated_at="2026-03-16T00:00:00+00:00",
             provider_runtime=schema.ProviderRuntime(
                 reasoning_provider="gemini",
-                planner_model="gemini-3.1-flash-lite-preview",
+                planner_model="gemini-3.1-flash-lite",
-                rerank_model="gemini-3.1-flash-lite-preview",
+                rerank_model="gemini-3.1-flash-lite",
             ),
             query_plan=schema.QueryPlan(
                 intent="comparison",
@@ -114,13 +114,13 @@ class CliV3Tests(unittest.TestCase):
     def test_slugify_and_emit_output_cover_supported_modes(self):
         report = self.make_report()
         self.assertEqual("openclaw-vs-nanoclaw", cli.slugify(report.topic))
-        self.assertEqual("last30days v3.0.0 CLI.", cli.__doc__)
+        self.assertEqual("last30days CLI.", cli.__doc__)
 
         compact = cli.emit_output(report, "compact")
         json_output = cli.emit_output(report, "json")
         context = cli.emit_output(report, "context")
 
-        self.assertIn("# last30days v3.0.0", compact)
+        self.assertIn("# last30days v", compact)
         self.assertIn('"topic": "OpenClaw vs NanoClaw"', json_output)
         self.assertIsInstance(context, str)
 

tests/test_env_cookies.py

@@ -114,9 +114,10 @@ class TestGetConfigCookieIntegration:
     @patch("lib.cookie_extract.extract_cookies")
     @patch("lib.env._find_project_env", return_value=None)
     @patch("lib.env.load_env_file", return_value={})
+    @patch("lib.env._load_keychain", return_value={})
     @patch("lib.env.get_openai_auth")
     def test_get_config_injects_cookies(
-        self, mock_openai, mock_load, mock_proj, mock_extract
+        self, mock_openai, mock_keychain, mock_load, mock_proj, mock_extract
     ):
         from lib.env import get_config, OpenAIAuth
         mock_openai.return_value = OpenAIAuth(

tests/test_env_keychain.py

@@ -0,0 +1,182 @@
+"""Tests for macOS Keychain credential source in lib/env.py.
+
+Covers:
+  - non-Darwin returns {}
+  - missing `security` binary returns {}
+  - successful lookups return parsed key/value pairs
+  - subprocess timeout / OSError are swallowed
+  - get_config merges keychain at lowest priority and labels _CONFIG_SOURCE
+"""
+
+from __future__ import annotations
+
+import re
+import subprocess
+import sys
+from pathlib import Path
+from unittest import mock
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "skills" / "last30days" / "scripts"))
+
+from lib import env  # noqa: E402
+
+SETUP_KEYCHAIN_SH = Path(__file__).resolve().parents[1] / "skills" / "last30days" / "scripts" / "setup-keychain.sh"
+
+
+# ---------------------------------------------------------------------------
+# _load_keychain unit tests
+# ---------------------------------------------------------------------------
+
+
+def test_load_keychain_returns_empty_on_non_darwin():
+    with mock.patch("platform.system", return_value="Linux"):
+        assert env._load_keychain(["XAI_API_KEY"]) == {}
+
+
+def test_load_keychain_returns_empty_when_security_missing():
+    with mock.patch("platform.system", return_value="Darwin"), \
+         mock.patch("shutil.which", return_value=None):
+        assert env._load_keychain(["XAI_API_KEY"]) == {}
+
+
+def _run_result(returncode: int, stdout: str = "") -> subprocess.CompletedProcess:
+    return subprocess.CompletedProcess(args=[], returncode=returncode, stdout=stdout, stderr="")
+
+
+def test_load_keychain_loads_present_keys_skips_missing():
+    def fake_run(cmd, **kwargs):
+        service = cmd[cmd.index("-s") + 1]
+        if service == "last30days-XAI_API_KEY":
+            return _run_result(0, "xai-abc\n")
+        if service == "last30days-BRAVE_API_KEY":
+            return _run_result(0, "brv-xyz\n")
+        return _run_result(44)  # security's "not found" exit code
+
+    with mock.patch("platform.system", return_value="Darwin"), \
+         mock.patch("shutil.which", return_value="/usr/bin/security"), \
+         mock.patch("subprocess.run", side_effect=fake_run):
+        result = env._load_keychain(["XAI_API_KEY", "BRAVE_API_KEY", "OPENAI_API_KEY"])
+
+    assert result == {"XAI_API_KEY": "xai-abc", "BRAVE_API_KEY": "brv-xyz"}
+
+
+def test_load_keychain_strips_whitespace_and_newlines():
+    with mock.patch("platform.system", return_value="Darwin"), \
+         mock.patch("shutil.which", return_value="/usr/bin/security"), \
+         mock.patch("subprocess.run", return_value=_run_result(0, "  hello-key  \n")):
+        result = env._load_keychain(["FOO"])
+    assert result == {"FOO": "hello-key"}
+
+
+def test_load_keychain_swallows_subprocess_errors():
+    def fake_run(cmd, **kwargs):
+        raise subprocess.TimeoutExpired(cmd=cmd, timeout=5)
+
+    with mock.patch("platform.system", return_value="Darwin"), \
+         mock.patch("shutil.which", return_value="/usr/bin/security"), \
+         mock.patch("subprocess.run", side_effect=fake_run):
+        assert env._load_keychain(["XAI_API_KEY"]) == {}
+
+
+def test_load_keychain_swallows_oserror():
+    with mock.patch("platform.system", return_value="Darwin"), \
+         mock.patch("shutil.which", return_value="/usr/bin/security"), \
+         mock.patch("subprocess.run", side_effect=OSError("boom")):
+        assert env._load_keychain(["XAI_API_KEY"]) == {}
+
+
+def test_load_keychain_skips_empty_stdout():
+    with mock.patch("platform.system", return_value="Darwin"), \
+         mock.patch("shutil.which", return_value="/usr/bin/security"), \
+         mock.patch("subprocess.run", return_value=_run_result(0, "")):
+        assert env._load_keychain(["XAI_API_KEY"]) == {}
+
+
+# ---------------------------------------------------------------------------
+# get_config integration tests
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture
+def clean_env(monkeypatch, tmp_path):
+    """Hide every key get_config might touch and point CONFIG_FILE at a
+    non-existent path so no real user config bleeds in."""
+    for var in [
+        "OPENAI_API_KEY", "XAI_API_KEY", "BRAVE_API_KEY", "AUTH_TOKEN", "CT0",
+        "SCRAPECREATORS_API_KEY", "APIFY_API_TOKEN", "BSKY_HANDLE",
+        "BSKY_APP_PASSWORD", "TRUTHSOCIAL_TOKEN", "EXA_API_KEY",
+        "SERPER_API_KEY", "OPENROUTER_API_KEY", "PARALLEL_API_KEY",
+        "XQUIK_API_KEY", "GOOGLE_API_KEY", "GEMINI_API_KEY",
+        "GOOGLE_GENAI_API_KEY", "INCLUDE_SOURCES", "FROM_BROWSER",
+    ]:
+        monkeypatch.delenv(var, raising=False)
+    monkeypatch.setattr(env, "CONFIG_FILE", tmp_path / "does-not-exist.env")
+    monkeypatch.chdir(tmp_path)  # no project .env in this tree either
+
+
+def test_get_config_reports_keychain_source(clean_env):
+    with mock.patch.object(env, "_load_keychain", return_value={"XAI_API_KEY": "xai-from-kc"}):
+        cfg = env.get_config()
+    assert cfg["_CONFIG_SOURCE"] == "keychain"
+    assert cfg["XAI_API_KEY"] == "xai-from-kc"
+
+
+def test_get_config_env_var_overrides_keychain(clean_env, monkeypatch):
+    monkeypatch.setenv("XAI_API_KEY", "xai-from-env")
+    with mock.patch.object(env, "_load_keychain", return_value={"XAI_API_KEY": "xai-from-kc"}):
+        cfg = env.get_config()
+    assert cfg["XAI_API_KEY"] == "xai-from-env"
+
+
+def test_get_config_reports_env_only_when_keychain_empty(clean_env):
+    with mock.patch.object(env, "_load_keychain", return_value={}):
+        cfg = env.get_config()
+    assert cfg["_CONFIG_SOURCE"] == "env_only"
+
+
+def test_get_config_global_file_outranks_keychain(clean_env, tmp_path, monkeypatch):
+    cfg_file = tmp_path / "global.env"
+    cfg_file.write_text("XAI_API_KEY=xai-from-file\n")
+    monkeypatch.setattr(env, "CONFIG_FILE", cfg_file)
+    with mock.patch.object(env, "_load_keychain", return_value={"XAI_API_KEY": "xai-from-kc"}):
+        cfg = env.get_config()
+    assert cfg["XAI_API_KEY"] == "xai-from-file"
+    assert cfg["_CONFIG_SOURCE"].startswith("global:")
+
+
+def test_get_config_openai_key_can_come_from_keychain(clean_env):
+    """OPENAI_API_KEY must be visible to get_openai_auth via the keychain
+    merge — wiring regression test."""
+    with mock.patch.object(env, "_load_keychain", return_value={"OPENAI_API_KEY": "sk-from-kc"}):
+        cfg = env.get_config()
+    assert cfg["OPENAI_API_KEY"] == "sk-from-kc"
+    assert cfg["OPENAI_AUTH_SOURCE"] == "api_key"
+
+
+# ---------------------------------------------------------------------------
+# Drift guard: lib/env.py KEYCHAIN_KEYS and setup-keychain.sh ALL_KEYS must
+# stay in lockstep. A mismatch means users storing a key via the helper script
+# wouldn't see it picked up by the loader, or vice versa.
+# ---------------------------------------------------------------------------
+
+
+def _parse_all_keys_from_shell(script: Path) -> list[str]:
+    text = script.read_text(encoding="utf-8")
+    match = re.search(r"ALL_KEYS=\(\s*(.*?)\s*\)", text, re.DOTALL)
+    if not match:
+        raise AssertionError(f"ALL_KEYS=( ... ) array not found in {script}")
+    body = match.group(1)
+    # Strip shell comments and split on whitespace
+    body = re.sub(r"#[^\n]*", "", body)
+    return [tok for tok in body.split() if tok]
+
+
+def test_keychain_keys_match_setup_script():
+    shell_keys = _parse_all_keys_from_shell(SETUP_KEYCHAIN_SH)
+    python_keys = list(env.KEYCHAIN_KEYS)
+    assert shell_keys == python_keys, (
+        "lib/env.py::KEYCHAIN_KEYS and scripts/setup-keychain.sh::ALL_KEYS "
+        f"have drifted.\n  python: {python_keys}\n  shell:  {shell_keys}"
+    )

tests/test_env_v3.py

@@ -41,6 +41,34 @@ class EnvV3Tests(unittest.TestCase):
             with mock.patch.dict(os.environ, {}, clear=False):
                 self.assertIsNone(bird_x.is_bird_authenticated())
 
+    def test_file_permission_check_skips_windows_posix_mode_bits(self):
+        path = mock.Mock(spec=Path)
+        with mock.patch.object(env.os, "name", "nt"), mock.patch.object(env.sys.stderr, "write") as write:
+            env._check_file_permissions(path)
+
+        path.stat.assert_not_called()
+        write.assert_not_called()
+
+
+class ThreadsAvailabilityTests(unittest.TestCase):
+    """Threads is in the SC default-on family: same key, same per-call cost
+    shape as TikTok / Instagram, so the same default-on rule applies.
+    Suppression goes through EXCLUDE_SOURCES, not gated opt-in."""
+
+    def test_threads_available_with_sc_key_only(self):
+        self.assertTrue(env.is_threads_available({"SCRAPECREATORS_API_KEY": "k"}))
+
+    def test_threads_unavailable_without_sc_key(self):
+        self.assertFalse(env.is_threads_available({}))
+        self.assertFalse(env.is_threads_available({"INCLUDE_SOURCES": "threads"}))
+
+    def test_threads_does_not_require_include_sources(self):
+        """Regression guard: INCLUDE_SOURCES should not be needed."""
+        self.assertTrue(env.is_threads_available({
+            "SCRAPECREATORS_API_KEY": "k",
+            "INCLUDE_SOURCES": "",
+        }))
+
 
 if __name__ == "__main__":
     unittest.main()

tests/test_evaluator_v3.py

@@ -125,7 +125,7 @@ class EvaluatorV3Tests(unittest.TestCase):
                 topic="test topic",
                 query_type="general",
                 items=[{"key": "a"}],
-                judge_model="gemini-3.1-flash-lite-preview",
+                judge_model="gemini-3.1-flash-lite",
                 gemini_api_key="key",
             )
             self.assertEqual({"a": 3}, cached)
@@ -136,7 +136,7 @@ class EvaluatorV3Tests(unittest.TestCase):
                 topic="test topic",
                 query_type="general",
                 items=[],
-                judge_model="gemini-3.1-flash-lite-preview",
+                judge_model="gemini-3.1-flash-lite",
                 gemini_api_key=None,
             )
             self.assertEqual({}, skipped)

tests/test_footer_nudge_suppression.py

@@ -6,6 +6,7 @@ from __future__ import annotations
 import os
 import subprocess
 import sys
+import tempfile
 import unittest
 from pathlib import Path
 
@@ -27,13 +28,31 @@ class FooterNudgeSuppressionTests(unittest.TestCase):
             "--emit=md",
             *argv,
         ]
-        env = {**os.environ, "LAST30DAYS_SKIP_PREFLIGHT": "1"}
+        env = {
+            **os.environ,
+            "LAST30DAYS_SKIP_PREFLIGHT": "1",
+            # Skip ~/.config/last30days/.env so a contributor's saved
+            # BRAVE/EXA/SERPER/PARALLEL key doesn't make grounding "available"
+            # and suppress the promo we're checking for.
+            "LAST30DAYS_CONFIG_DIR": "",
+            # Pin X as available so _missing_sources_for_promo selects "web"
+            # (otherwise the "x" promo wins and the BRAVE_API_KEY string never
+            # appears).
+            "XAI_API_KEY": "test-stub",
+        }
         # Strip any grounded-web keys the host might have so the promo path
-        # triggers deterministically in mock + no-backend.
+        # triggers deterministically in mock + no-backend. Also strip X cookie
+        # credentials so XAI_API_KEY is the unambiguous X backend.
         for key in ("BRAVE_API_KEY", "EXA_API_KEY", "SERPER_API_KEY",
-                    "PARALLEL_API_KEY", "OPENROUTER_API_KEY"):
+                    "PARALLEL_API_KEY", "OPENROUTER_API_KEY",
+                    "AUTH_TOKEN", "CT0", "LAST30DAYS_X_BACKEND"):
             env.pop(key, None)
-        return subprocess.run(cmd, capture_output=True, text=True, env=env)
+        # Run from a tmpdir so _find_project_env() can't walk up into any
+        # .claude/last30days.env above the repo on the contributor's machine.
+        with tempfile.TemporaryDirectory() as tmp:
+            return subprocess.run(
+                cmd, capture_output=True, text=True, env=env, cwd=tmp,
+            )
 
     def test_bare_run_emits_web_promo(self):
         result = self._run(topic="OpenAI")

tests/test_grounding_v3.py

@@ -190,5 +190,88 @@ class WebSearchDispatchTests(unittest.TestCase):
             grounding.web_search("test", ("2026-02-25", "2026-03-27"), {}, backend="google")
 
 
+class RedditEnrichmentGateTests(unittest.TestCase):
+    """EXCLUDE_SOURCES=reddit must suppress the web-search Reddit enrichment.
+
+    Otherwise a user who explicitly excluded Reddit would still get Reddit
+    content smuggled back in via web-search URLs that happen to point at
+    reddit.com threads.
+    """
+
+    def test_reddit_excluded_via_exclude_sources_skips_enrichment(self):
+        config = {"BRAVE_API_KEY": "k", "EXCLUDE_SOURCES": "reddit"}
+        items = [{"url": "https://www.reddit.com/r/python/comments/abc/title/", "snippet": "original"}]
+        with patch("lib.grounding.brave_search", return_value=(items, {})), \
+             patch("lib.grounding._enrich_reddit_items") as enrich_mock:
+            grounding.web_search("test", ("2026-02-25", "2026-03-27"), config, backend="auto")
+            enrich_mock.assert_not_called()
+
+    def test_reddit_excluded_case_insensitive(self):
+        for value in ("REDDIT", "Reddit", " reddit ", "x,reddit,y"):
+            config = {"BRAVE_API_KEY": "k", "EXCLUDE_SOURCES": value}
+            self.assertTrue(
+                grounding._reddit_excluded(config),
+                msg=f"_reddit_excluded should be True for EXCLUDE_SOURCES={value!r}",
+            )
+
+    def test_reddit_not_excluded_when_other_sources_listed(self):
+        config = {"EXCLUDE_SOURCES": "tiktok,instagram"}
+        self.assertFalse(grounding._reddit_excluded(config))
+
+    def test_enrichment_runs_when_reddit_not_excluded(self):
+        config = {"BRAVE_API_KEY": "k"}
+        items = [{"url": "https://www.reddit.com/r/python/comments/abc/title/", "snippet": "original"}]
+        with patch("lib.grounding.brave_search", return_value=(items, {})), \
+             patch("lib.grounding._enrich_reddit_items", return_value=items) as enrich_mock:
+            grounding.web_search("test", ("2026-02-25", "2026-03-27"), config, backend="auto")
+            enrich_mock.assert_called_once()
+
+
+class RedditEnrichItemsTests(unittest.TestCase):
+    """Direct tests for `_enrich_reddit_items` covering the selftext key path
+    and the RedditRateLimitError early-exit behavior.
+    """
+
+    def test_selftext_under_submission_populates_snippet(self):
+        from lib import reddit_enrich
+
+        item = {
+            "url": "https://www.reddit.com/r/python/comments/abc/title/",
+            "snippet": "original",
+        }
+        parsed = {
+            "submission": {"selftext": "thread body content"},
+            "comments": [],
+        }
+        with patch.object(reddit_enrich, "fetch_thread_data", return_value={"raw": True}), \
+             patch.object(reddit_enrich, "parse_thread_data", return_value=parsed):
+            result = grounding._enrich_reddit_items([item])
+        self.assertEqual("thread body content", result[0]["snippet"])
+        self.assertEqual("reddit_json_api", result[0]["enriched_via"])
+
+    def test_rate_limit_error_halts_iteration(self):
+        from lib import reddit_enrich
+
+        item1 = {"url": "https://www.reddit.com/r/python/comments/aaa/x/"}
+        item2 = {"url": "https://www.reddit.com/r/python/comments/bbb/y/"}
+
+        def fake_fetch(url, *args, **kwargs):
+            raise reddit_enrich.RedditRateLimitError(f"429 for {url}")
+
+        captured_stderr: list[str] = []
+
+        with patch.object(reddit_enrich, "fetch_thread_data", side_effect=fake_fetch) as fetch_mock, \
+             patch("lib.grounding.sys.stderr.write", side_effect=lambda s: captured_stderr.append(s)):
+            grounding._enrich_reddit_items([item1, item2])
+
+        # Only the first item should have triggered a fetch attempt
+        self.assertEqual(1, fetch_mock.call_count)
+        # A stderr message about the rate-limit halt should have been emitted
+        self.assertTrue(
+            any("rate-limited" in msg.lower() or "rate limited" in msg.lower() for msg in captured_stderr),
+            msg=f"Expected a rate-limit stderr message, got: {captured_stderr!r}",
+        )
+
+
 if __name__ == "__main__":
     unittest.main()

tests/test_hackernews.py

@@ -160,12 +160,44 @@ def test_title_matches_query_empty_query():
 
 
 def test_title_matches_query_partial_match():
-    """Test that all query words must match."""
+    """Any-word matching: at least one query token in title is enough.
+
+    Previously required *all* tokens, which killed every hit on multi-keyword
+    theme queries like 'claude, personal agents, agentic infra' since no real
+    HN title contains all 5 tokens verbatim. Token-overlap relevance at parse
+    time still demotes weak matches, so the loosened gate is safe.
+    """
     title = "New AI framework"
     query = "AI blockchain"
-    
+
-    # "blockchain" is not in title, so should fail
+    # "AI" matches as a whole word, even though "blockchain" doesn't appear
-    assert hackernews._title_matches_query(title, query) is False
+    assert hackernews._title_matches_query(title, query) is True
+
+
+def test_title_matches_query_no_token_in_title():
+    """If no query token appears in the title at all, reject."""
+    assert hackernews._title_matches_query("New rust compiler", "AI blockchain") is False
+
+
+def test_title_matches_query_word_boundary_not_substring():
+    """Short tokens must match on word boundaries, not as substrings.
+
+    Without word-boundary matching, 'ai' would falsely match 'email',
+    'rail', 'artists', etc.
+    """
+    # 'ai' as a substring of 'email' must not match
+    assert hackernews._title_matches_query("New email service", "ai blockchain") is False
+    # 'ai' as a whole word does match
+    assert hackernews._title_matches_query("Cool AI tool launched", "ai blockchain") is True
+
+
+def test_title_matches_query_flattens_hyphens_and_commas():
+    """Query tokens split on hyphens/commas the same way search_hackernews
+    flattens them, so the post-filter stays aligned with what Algolia saw."""
+    # query 'ts-bun-node' flattens to ['ts', 'bun', 'node']; title contains 'bun'
+    assert hackernews._title_matches_query("Bun 1.2 released", "ts-bun-node") is True
+    # query 'rust, go, zig' flattens; title contains 'go'
+    assert hackernews._title_matches_query("Go 1.24 generics update", "rust, go, zig") is True
 
 
 # === Tests for search_hackernews() ===

tests/test_html_render.py

@@ -277,6 +277,26 @@ class HtmlCliIntegrationTests(unittest.TestCase):
             path = cli.compute_save_path_display("/tmp", report.topic, "v3", "html")
             self.assertTrue(path.endswith("/ai-agent-frameworks-raw-html-v3.html"))
 
+    def test_save_output_can_persist_comparison_html(self):
+        reports = [
+            ("OpenClaw", _report("OpenClaw", ["Containers"])),
+            ("Hermes", _report("Hermes", ["Memory"])),
+        ]
+        rendered = cli.emit_comparison_output(reports, "html")
+        with tempfile.TemporaryDirectory() as tmpdir:
+            path = cli.save_output(
+                reports[0][1],
+                "html",
+                tmpdir,
+                topic_override=cli.comparison_topic(reports),
+                rendered_content=rendered,
+            )
+            self.assertEqual("openclaw-vs-hermes-raw-html.html", path.name)
+            saved = path.read_text(encoding="utf-8")
+        self.assertIn("last30days · OpenClaw vs Hermes", saved)
+        self.assertIn("comparing 2: OpenClaw, Hermes", saved)
+        self.assertNotIn("last30days · OpenClaw</title>", saved)
+
 
 if __name__ == "__main__":
     unittest.main()

tests/test_http_v3.py

@@ -104,3 +104,117 @@ class TestParamsEncoding(unittest.TestCase):
         sent_url = self._sent_url(mock_urlopen)
         self.assertIn("count=25", sent_url)
         self.assertIn("raw=True", sent_url)
+
+
+class TestDNSResolutionRetry(unittest.TestCase):
+    """DNS resolution failures (gaierror) must retry with exponential backoff.
+
+    Caller-passed `retries` values smaller than MIN_DNS_RETRIES are expanded
+    on the first gaierror so a transient resolution failure doesn't wipe a
+    request just because the caller passed retries=2.
+    """
+
+    @patch("lib.http.urllib.request.urlopen")
+    @patch("lib.http.time.sleep")
+    def test_gaierror_retries_up_to_min_dns_retries_even_when_caller_passes_fewer(
+        self, mock_sleep, mock_urlopen
+    ):
+        """Caller passed retries=2; gaierror should still get MIN_DNS_RETRIES attempts."""
+        import socket
+        err = urllib.error.URLError(socket.gaierror(-2, "Name or service not known"))
+        mock_urlopen.side_effect = err
+
+        with self.assertRaises(http.HTTPError):
+            http.request("GET", "http://nonexistent.example", retries=2)
+
+        # Caller passed retries=2, but the budget expanded to MIN_DNS_RETRIES=3.
+        self.assertEqual(mock_urlopen.call_count, http.MIN_DNS_RETRIES)
+
+    @patch("lib.http.urllib.request.urlopen")
+    @patch("lib.http.time.sleep")
+    def test_gaierror_succeeds_after_transient_failure(self, mock_sleep, mock_urlopen):
+        """gaierror on attempt 1, then success — should NOT raise."""
+        import socket
+        success_response = MagicMock()
+        success_response.read.return_value = b'{"ok": true}'
+        success_response.status = 200
+        success_response.__enter__ = lambda self: self
+        success_response.__exit__ = lambda *args: None
+
+        err = urllib.error.URLError(socket.gaierror(-2, "Name or service not known"))
+        mock_urlopen.side_effect = [err, success_response]
+
+        result = http.request("GET", "http://flaky.example", retries=2)
+
+        self.assertEqual(result, {"ok": True})
+        self.assertEqual(mock_urlopen.call_count, 2)
+
+    @patch("lib.http.urllib.request.urlopen")
+    @patch("lib.http.time.sleep")
+    def test_gaierror_uses_exponential_backoff(self, mock_sleep, mock_urlopen):
+        """Backoff delays for gaierror should be 1s, 2s, 4s — not the linear default."""
+        import socket
+        err = urllib.error.URLError(socket.gaierror(-2, "Name or service not known"))
+        mock_urlopen.side_effect = err
+
+        with self.assertRaises(http.HTTPError):
+            http.request("GET", "http://nonexistent.example", retries=3)
+
+        # Expected sleep calls: 1s (after attempt 1), 2s (after attempt 2).
+        # No sleep after the final attempt (the loop exits to raise).
+        sleep_delays = [call.args[0] for call in mock_sleep.call_args_list]
+        self.assertEqual(sleep_delays, [1, 2])
+
+    @patch("lib.http.urllib.request.urlopen")
+    @patch("lib.http.time.sleep")
+    def test_non_dns_urlerror_uses_linear_backoff_not_dns_branch(
+        self, mock_sleep, mock_urlopen
+    ):
+        """A URLError that's NOT a gaierror must NOT expand the retry budget."""
+        # ConnectionRefusedError-style URLError reason (not gaierror)
+        err = urllib.error.URLError(ConnectionRefusedError(111, "Connection refused"))
+        mock_urlopen.side_effect = err
+
+        with self.assertRaises(http.HTTPError):
+            http.request("GET", "http://refused.example", retries=2)
+
+        # Caller passed retries=2, and non-DNS URLError doesn't expand it.
+        self.assertEqual(mock_urlopen.call_count, 2)
+
+    @patch("lib.http.urllib.request.urlopen")
+    @patch("lib.http.time.sleep")
+    def test_dns_widening_does_not_leak_into_subsequent_non_dns_urlerror(
+        self, mock_sleep, mock_urlopen
+    ):
+        """Mixed sequence: DNS-then-non-DNS must respect caller's original retries.
+
+        Without the fix, the first gaierror widens effective_retries from 2 to
+        MIN_DNS_RETRIES=3, and a subsequent ConnectionRefused on attempt 1
+        slips into a third overall attempt — exceeding what the caller asked
+        for. Each non-DNS error path must gate on the original `retries`.
+        """
+        import socket
+        dns_err = urllib.error.URLError(socket.gaierror(-2, "Name or service not known"))
+        conn_err = urllib.error.URLError(ConnectionRefusedError(111, "Connection refused"))
+        mock_urlopen.side_effect = [dns_err, conn_err, conn_err]  # 3rd would only fire if budget leaked
+
+        with self.assertRaises(http.HTTPError):
+            http.request("GET", "http://flaky.example", retries=2)
+
+        # Caller asked for at most 2 attempts. DNS widening must not give us a 3rd.
+        self.assertEqual(mock_urlopen.call_count, 2)
+
+    @patch("lib.http.urllib.request.urlopen")
+    @patch("lib.http.time.sleep")
+    def test_dns_widening_does_not_leak_into_subsequent_oserror(
+        self, mock_sleep, mock_urlopen
+    ):
+        """Mixed sequence: DNS-then-OSError must respect caller's original retries."""
+        import socket
+        dns_err = urllib.error.URLError(socket.gaierror(-2, "Name or service not known"))
+        mock_urlopen.side_effect = [dns_err, TimeoutError("timed out"), TimeoutError("timed out")]
+
+        with self.assertRaises(http.HTTPError):
+            http.request("GET", "http://flaky.example", retries=2)
+
+        self.assertEqual(mock_urlopen.call_count, 2)

tests/test_pipeline_v3.py

@@ -904,5 +904,81 @@ class TestZeroKeyPipelineRun(unittest.TestCase):
             self.assertEqual("fallback-local-score", candidate.explanation)
 
 
+class TestExcludeSources(unittest.TestCase):
+    """EXCLUDE_SOURCES env var filters sources out of available_sources().
+
+    The existing INCLUDE_SOURCES allowlist (used by Perplexity opt-in) does
+    not cover this case — tiktok and instagram are added unconditionally
+    when SCRAPECREATORS_API_KEY is set, with no way to opt out short of
+    unsetting the key. EXCLUDE_SOURCES gives runs a per-invocation denylist.
+    """
+
+    def test_excludes_tiktok_and_instagram(self):
+        config = {
+            "SCRAPECREATORS_API_KEY": "test-key",
+            "EXCLUDE_SOURCES": "tiktok,instagram",
+        }
+        sources = pipeline.available_sources(config)
+        self.assertNotIn("tiktok", sources)
+        self.assertNotIn("instagram", sources)
+        self.assertIn("reddit", sources)
+        self.assertIn("hackernews", sources)
+
+    def test_no_exclusion_when_unset(self):
+        config = {"SCRAPECREATORS_API_KEY": "test-key"}
+        sources = pipeline.available_sources(config)
+        self.assertIn("tiktok", sources)
+        self.assertIn("instagram", sources)
+
+    def test_empty_exclude_sources_is_noop(self):
+        config = {
+            "SCRAPECREATORS_API_KEY": "test-key",
+            "EXCLUDE_SOURCES": "",
+        }
+        sources = pipeline.available_sources(config)
+        self.assertIn("tiktok", sources)
+        self.assertIn("instagram", sources)
+
+    def test_whitespace_and_case_insensitive(self):
+        config = {
+            "SCRAPECREATORS_API_KEY": "test-key",
+            "EXCLUDE_SOURCES": " TikTok , INSTAGRAM ",
+        }
+        sources = pipeline.available_sources(config)
+        self.assertNotIn("tiktok", sources)
+        self.assertNotIn("instagram", sources)
+
+    def test_excludes_non_scrapecreators_source(self):
+        """EXCLUDE_SOURCES applies to any source, not just SC-backed ones."""
+        config = {"EXCLUDE_SOURCES": "hackernews"}
+        sources = pipeline.available_sources(config)
+        self.assertNotIn("hackernews", sources)
+        self.assertIn("reddit", sources)
+
+
+class TestExcludeSourcesEndToEnd(unittest.TestCase):
+    """Wiring regression: EXCLUDE_SOURCES from the process environment must
+    reach available_sources() via env.get_config(). The unit tests above
+    construct config dicts directly; this one exercises the env-to-config
+    path so a missing entry in env.py's keys list is caught immediately."""
+
+    def test_exclude_sources_from_env_propagates_through_get_config(self):
+        import os
+        from unittest.mock import patch as _patch
+        from lib import env as env_mod
+        from importlib import reload
+        with _patch.dict(os.environ, {
+            "LAST30DAYS_CONFIG_DIR": "",
+            "EXCLUDE_SOURCES": "tiktok,instagram",
+            "SCRAPECREATORS_API_KEY": "fake",
+        }, clear=False):
+            reload(env_mod)
+            cfg = env_mod.get_config()
+        self.assertEqual(cfg.get("EXCLUDE_SOURCES"), "tiktok,instagram")
+        sources = pipeline.available_sources(cfg)
+        self.assertNotIn("tiktok", sources)
+        self.assertNotIn("instagram", sources)
+
+
 if __name__ == "__main__":
     unittest.main()

tests/test_plugin_contract.py

@@ -1,5 +1,5 @@
 import json
-import re
+import sys
 import tomllib
 import unittest
 from pathlib import Path
@@ -8,17 +8,19 @@ from pathlib import Path
 ROOT = Path(__file__).resolve().parents[1]
 SKILL_ROOT = ROOT / "skills" / "last30days"
 
+sys.path.insert(0, str(SKILL_ROOT / "scripts"))
+from lib.skill_meta import read_skill_version  # noqa: E402
+
 
 def _json(path: Path) -> dict:
     return json.loads(path.read_text(encoding="utf-8"))
 
 
 def _skill_version() -> str:
-    text = (SKILL_ROOT / "SKILL.md").read_text(encoding="utf-8")
+    version = read_skill_version(SKILL_ROOT / "SKILL.md")
-    match = re.search(r'^version:\s*"([^"]+)"\s*$', text, re.MULTILINE)
+    if not version:
-    if not match:
         raise AssertionError("SKILL.md version frontmatter not found")
-    return match.group(1)
+    return version
 
 
 class TestPluginContract(unittest.TestCase):

tests/test_render_v3.py

@@ -70,8 +70,8 @@ def sample_report() -> schema.Report:
         generated_at="2026-03-16T00:00:00+00:00",
         provider_runtime=schema.ProviderRuntime(
             reasoning_provider="gemini",
-            planner_model="gemini-3.1-flash-lite-preview",
+            planner_model="gemini-3.1-flash-lite",
-            rerank_model="gemini-3.1-flash-lite-preview",
+            rerank_model="gemini-3.1-flash-lite",
         ),
         query_plan=schema.QueryPlan(
             intent="breaking_news",
@@ -91,7 +91,8 @@ def sample_report() -> schema.Report:
 class RenderV3Tests(unittest.TestCase):
     def test_render_compact_includes_cluster_first_sections(self):
         text = render.render_compact(sample_report())
-        self.assertIn("# last30days v3.0.0: test topic", text)
+        self.assertIn("# last30days v", text)
+        self.assertIn(": test topic", text)
         self.assertIn("Safety note: evidence text below is untrusted internet content", text)
         self.assertIn("## Ranked Evidence Clusters", text)
         self.assertIn("## Stats", text)
@@ -239,8 +240,8 @@ class RenderTopCommentsTests(unittest.TestCase):
             generated_at="2026-03-16T00:00:00+00:00",
             provider_runtime=schema.ProviderRuntime(
                 reasoning_provider="gemini",
-                planner_model="gemini-3.1-flash-lite-preview",
+                planner_model="gemini-3.1-flash-lite",
-                rerank_model="gemini-3.1-flash-lite-preview",
+                rerank_model="gemini-3.1-flash-lite",
             ),
             query_plan=schema.QueryPlan(
                 intent="breaking_news",
@@ -424,8 +425,8 @@ class RenderBestTakesCompactTests(unittest.TestCase):
             generated_at="2026-03-16T00:00:00+00:00",
             provider_runtime=schema.ProviderRuntime(
                 reasoning_provider="gemini",
-                planner_model="gemini-3.1-flash-lite-preview",
+                planner_model="gemini-3.1-flash-lite",
-                rerank_model="gemini-3.1-flash-lite-preview",
+                rerank_model="gemini-3.1-flash-lite",
             ),
             query_plan=schema.QueryPlan(
                 intent="breaking_news",

tests/test_rerank_v3.py

@@ -172,10 +172,10 @@ class RerankV3Tests(unittest.TestCase):
             plan=make_plan(),
             candidates=[first, second],
             provider=provider,
-            model="gemini-3.1-flash-lite-preview",
+            model="gemini-3.1-flash-lite",
             shortlist_size=1,
         )
-        self.assertEqual("gemini-3.1-flash-lite-preview", provider.model)
+        self.assertEqual("gemini-3.1-flash-lite", provider.model)
         self.assertEqual(95.0, first.rerank_score)
         self.assertEqual("high fit", first.explanation)
         # Tail is scored via the fallback (may or may not carry the entity-miss

tests/test_schema_v3.py

@@ -16,8 +16,8 @@ class SchemaV3Tests(unittest.TestCase):
             generated_at="2026-03-16T00:00:00+00:00",
             provider_runtime=schema.ProviderRuntime(
                 reasoning_provider="gemini",
-                planner_model="gemini-3.1-flash-lite-preview",
+                planner_model="gemini-3.1-flash-lite",
-                rerank_model="gemini-3.1-flash-lite-preview",
+                rerank_model="gemini-3.1-flash-lite",
             ),
             query_plan=schema.QueryPlan(
                 intent="breaking_news",

tests/test_security_workflow.py

@@ -0,0 +1,53 @@
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+WORKFLOW = ROOT / ".github" / "workflows" / "security.yml"
+# AGENTS.md is the canonical agent-guidance file; CLAUDE.md is a one-line
+# pointer (`@AGENTS.md`) so anything Claude Code-shaped reads the same source.
+AGENTS = ROOT / "AGENTS.md"
+
+
+def _workflow_text() -> str:
+    return WORKFLOW.read_text(encoding="utf-8")
+
+
+def test_security_workflow_exists() -> None:
+    assert WORKFLOW.is_file()
+
+
+def test_security_workflow_runs_dependency_audit_advisory_first() -> None:
+    text = _workflow_text()
+
+    assert "dependency-audit:" in text
+    assert "pip-audit" in text
+    assert "continue-on-error: true" in text
+    assert "Set continue-on-error: false once a clean baseline run is confirmed" in text
+
+
+def test_security_workflow_runs_secret_scan_for_pull_requests_and_main_pushes() -> None:
+    text = _workflow_text()
+
+    assert "secret-scan:" in text
+    assert "trufflesecurity/trufflehog" in text
+    assert "github.event_name == 'pull_request'" in text
+    assert "github.event_name == 'push'" in text
+    assert "--only-verified" in text
+
+
+def test_security_workflow_documents_advisory_policy() -> None:
+    text = _workflow_text()
+
+    assert "advisory-first" in text.lower()
+    assert "does not block merges" in text.lower()
+    assert "fixtures" in text.lower()
+    assert "env-based auth" in text.lower()
+
+
+def test_agent_guidance_mentions_secret_hygiene() -> None:
+    text = AGENTS.read_text(encoding="utf-8")
+
+    assert "Security hygiene" in text
+    assert "Never commit real API keys" in text
+    assert "skills/last30days/scripts/lib/env.py" in text
+    assert "fixtures" in text

tests/test_setup_openclaw.py

@@ -64,6 +64,19 @@ class TestRunOpenclawSetup:
         assert result["keys"]["brave"] is True
         assert result["keys"]["scrapecreators"] is False
 
+    def test_openclaw_metadata_keeps_scrapecreators_optional(self):
+        """OpenClaw metadata should not hard-require the ScrapeCreators key."""
+        skill_md = Path(__file__).parent.parent / "skills" / "last30days" / "SKILL.md"
+        text = skill_md.read_text()
+        assert "SCRAPECREATORS_API_KEY" in text
+        expected = (
+            "requires:\n"
+            "      env: []\n"
+            "      optionalEnv:\n"
+            "        - SCRAPECREATORS_API_KEY"
+        )
+        assert expected in text
+
     @patch("shutil.which")
     def test_x_method_xai(self, mock_which):
         """x_method is 'xai' when XAI_API_KEY is set."""
@@ -200,7 +213,9 @@ class TestPollDeviceAuth:
     @patch("lib.setup_wizard.urlopen")
     def test_timeout_returns_none(self, mock_urlopen, mock_time):
         """Returns None when timeout is exceeded."""
-        # Simulate time passing beyond deadline
+        # poll_device_auth captures started_at once, derives deadline + last_reminder
+        # from it, then checks time.time() in the while-loop. Two values: started_at,
+        # then a value past the deadline so the loop exits immediately.
         mock_time.time = MagicMock(side_effect=[0, 301])
         mock_time.sleep = MagicMock()
 
@@ -211,7 +226,9 @@ class TestPollDeviceAuth:
     @patch("lib.setup_wizard.urlopen")
     def test_expired_token_returns_none(self, mock_urlopen, mock_time):
         """Returns None on expired_token error."""
-        mock_time.time = MagicMock(side_effect=[0, 0])
+        # Loop terminates via urlopen response, not the clock — pin time to 0
+        # so the deadline check stays a non-event regardless of call count.
+        mock_time.time = MagicMock(return_value=0)
         mock_time.sleep = MagicMock()
 
         expired_resp = MagicMock()
@@ -230,7 +247,7 @@ class TestPollDeviceAuth:
         """HTTP 400 during polling continues (authorization pending)."""
         from urllib.error import HTTPError
 
-        mock_time.time = MagicMock(side_effect=[0, 0, 0])
+        mock_time.time = MagicMock(return_value=0)
         mock_time.sleep = MagicMock()
 
         success_resp = MagicMock()

tests/test_skill_meta.py

@@ -0,0 +1,61 @@
+"""Direct unit tests for skill_meta.read_skill_version.
+
+Covers the helper's own contract independent of render._skill_version which
+exercises it transitively. Without these, regressions in error handling or
+regex coverage inside the helper could pass CI because render.py's fallback
+to "?" swallows the signal.
+"""
+
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT / "skills" / "last30days" / "scripts"))
+from lib.skill_meta import read_skill_version  # noqa: E402
+
+
+class ReadSkillVersionTests(unittest.TestCase):
+    def setUp(self) -> None:
+        self._tmp = tempfile.TemporaryDirectory()
+        self.tmp_path = Path(self._tmp.name)
+
+    def tearDown(self) -> None:
+        self._tmp.cleanup()
+
+    def _write_skill_md(self, body: str) -> Path:
+        path = self.tmp_path / "SKILL.md"
+        path.write_text(body)
+        return path
+
+    def test_double_quoted_version(self) -> None:
+        path = self._write_skill_md('---\nname: x\nversion: "9.9.9"\n---\n')
+        self.assertEqual("9.9.9", read_skill_version(path))
+
+    def test_single_quoted_version(self) -> None:
+        path = self._write_skill_md("---\nname: x\nversion: '8.8.8'\n---\n")
+        self.assertEqual("8.8.8", read_skill_version(path))
+
+    def test_unquoted_version(self) -> None:
+        path = self._write_skill_md("---\nname: x\nversion: 7.7.7\n---\n")
+        self.assertEqual("7.7.7", read_skill_version(path))
+
+    def test_missing_file_returns_none(self) -> None:
+        self.assertIsNone(read_skill_version(self.tmp_path / "does-not-exist.md"))
+
+    def test_no_version_line_returns_none(self) -> None:
+        path = self._write_skill_md("---\nname: x\n---\n# body without version\n")
+        self.assertIsNone(read_skill_version(path))
+
+    def test_undecodable_bytes_returns_none(self) -> None:
+        # Bytes 128-255 don't form valid UTF-8 sequences; read_text() raises
+        # UnicodeDecodeError which the helper must catch.
+        path = self.tmp_path / "SKILL.md"
+        path.write_bytes(bytes(range(128, 256)))
+        self.assertIsNone(read_skill_version(path))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_store.py

@@ -3,7 +3,7 @@
 import json
 import sqlite3
 import tempfile
-from datetime import datetime, timedelta
+from datetime import datetime, timedelta, timezone
 from pathlib import Path
 
 import pytest
@@ -59,7 +59,68 @@ def sample_report():
             "source_weights": {},
         },
         "clusters": [],
-        "ranked_candidates": [],
+        "ranked_candidates": [
+            {
+                "candidate_id": "c-r1",
+                "item_id": "R1",
+                "source": "reddit",
+                "title": "Test Reddit Post",
+                "url": "https://reddit.com/r/test/1",
+                "snippet": "Reddit snippet",
+                "subquery_labels": ["primary"],
+                "native_ranks": {"reddit": 1},
+                "local_relevance": 0.8,
+                "freshness": 100,
+                "engagement": 50.0,
+                "source_quality": 0.8,
+                "rrf_score": 1.0,
+                "final_score": 0.8,
+                "explanation": "Reddit snippet",
+                "source_items": [
+                    {
+                        "item_id": "R1",
+                        "source": "reddit",
+                        "title": "Test Reddit Post",
+                        "body": "Reddit discussion content",
+                        "url": "https://reddit.com/r/test/1",
+                        "author": "testuser",
+                        "engagement_score": 50.0,
+                        "local_relevance": 0.8,
+                        "snippet": "Reddit snippet",
+                    }
+                ],
+            },
+            {
+                "candidate_id": "c-x1",
+                "item_id": "X1",
+                "source": "x",
+                "title": "Test X Post",
+                "url": "https://x.com/test/status/1",
+                "snippet": "X snippet",
+                "subquery_labels": ["primary"],
+                "native_ranks": {"x": 1},
+                "local_relevance": 0.85,
+                "freshness": 100,
+                "engagement": 75.0,
+                "source_quality": 0.8,
+                "rrf_score": 1.0,
+                "final_score": 0.85,
+                "explanation": "X snippet",
+                "source_items": [
+                    {
+                        "item_id": "X1",
+                        "source": "x",
+                        "title": "Test X Post",
+                        "body": "X post content",
+                        "url": "https://x.com/test/status/1",
+                        "author": "xuser",
+                        "engagement_score": 75.0,
+                        "local_relevance": 0.85,
+                        "snippet": "X snippet",
+                    }
+                ],
+            },
+        ],
         "items_by_source": {
             "reddit": [
                 {
@@ -236,13 +297,13 @@ def test_findings_from_report_handles_missing_fields():
         "clusters": [],
         "ranked_candidates": [],
         "items_by_source": {
-            "reddit": [
+            "hackernews": [
                 {
                     "item_id": "R1",
-                    "source": "reddit",
+                    "source": "hackernews",
                     "title": "Test",
                     "body": "Content",
-                    "url": "https://reddit.com/1",
+                    "url": "https://news.ycombinator.com/item?id=1",
                     "author": None,  # Missing author
                     "engagement_score": None,  # Missing engagement
                     "local_relevance": None,  # Missing relevance
@@ -384,6 +445,127 @@ def test_store_findings_skips_items_without_url(temp_db):
     assert counts["new"] == 1
 
 
+def test_init_db_creates_finding_sightings_table(temp_db):
+    """Test that the per-run sightings ledger is available on fresh databases."""
+    conn = sqlite3.connect(str(temp_db))
+    table = conn.execute(
+        "SELECT name FROM sqlite_master WHERE type='table' AND name='finding_sightings'"
+    ).fetchone()
+    columns = {
+        row[1]: row[3]
+        for row in conn.execute("PRAGMA table_info(finding_sightings)").fetchall()
+    }
+    conn.close()
+
+    assert table is not None
+    assert columns["finding_id"] == 1
+
+
+def test_store_findings_records_sightings_for_new_findings(temp_db):
+    """Test that each stored finding is linked to the run that observed it."""
+    topic = store.add_topic("Test Topic")
+    run_id = store.record_run(topic["id"], source_mode="v3")
+    findings = [
+        {
+            "source": "reddit",
+            "source_url": "https://reddit.com/1",
+            "source_title": "Reddit 1",
+            "content": "Content 1",
+            "engagement_score": 10.0,
+            "relevance_score": 0.7,
+        },
+        {
+            "source": "x",
+            "source_url": "https://x.com/a/status/1",
+            "source_title": "X 1",
+            "content": "Content 2",
+            "engagement_score": 20.0,
+            "relevance_score": 0.8,
+        },
+    ]
+
+    store.store_findings(run_id, topic["id"], findings)
+
+    sightings = store.get_sightings_for_run(topic["id"], run_id)
+    assert [s["source_url"] for s in sightings] == [
+        "https://reddit.com/1",
+        "https://x.com/a/status/1",
+    ]
+    assert {s["source"] for s in sightings} == {"reddit", "x"}
+
+
+def test_store_findings_records_sightings_for_resighted_findings(temp_db):
+    """Test that a re-seen finding is recorded for each run that observes it."""
+    topic = store.add_topic("Test Topic")
+    first_run_id = store.record_run(topic["id"], source_mode="v3")
+    second_run_id = store.record_run(topic["id"], source_mode="v3")
+    finding = {
+        "source": "reddit",
+        "source_url": "https://reddit.com/1",
+        "source_title": "Reddit 1",
+        "content": "Content",
+        "engagement_score": 10.0,
+        "relevance_score": 0.7,
+    }
+
+    store.store_findings(first_run_id, topic["id"], [finding])
+    store.store_findings(second_run_id, topic["id"], [{**finding, "engagement_score": 15.0}])
+
+    first_sightings = store.get_sightings_for_run(topic["id"], first_run_id)
+    second_sightings = store.get_sightings_for_run(topic["id"], second_run_id)
+
+    assert len(first_sightings) == 1
+    assert len(second_sightings) == 1
+    assert first_sightings[0]["source_url"] == second_sightings[0]["source_url"]
+    assert second_sightings[0]["engagement_score"] == 15.0
+
+
+def test_store_findings_sightings_are_idempotent_per_run(temp_db):
+    """Test that storing the same finding twice for one run does not duplicate sightings."""
+    topic = store.add_topic("Test Topic")
+    run_id = store.record_run(topic["id"], source_mode="v3")
+    finding = {
+        "source": "reddit",
+        "source_url": "https://reddit.com/1",
+        "source_title": "Reddit 1",
+        "content": "Content",
+        "engagement_score": 10.0,
+        "relevance_score": 0.7,
+    }
+
+    store.store_findings(run_id, topic["id"], [finding])
+    store.store_findings(run_id, topic["id"], [finding])
+
+    sightings = store.get_sightings_for_run(topic["id"], run_id)
+    assert len(sightings) == 1
+
+
+def test_store_findings_updates_existing_sighting_for_same_run(temp_db):
+    """Test that retrying a run refreshes its sighting snapshot instead of freezing it."""
+    topic = store.add_topic("Test Topic")
+    run_id = store.record_run(topic["id"], source_mode="v3")
+    finding = {
+        "source": "reddit",
+        "source_url": "https://reddit.com/1",
+        "source_title": "Reddit 1",
+        "content": "Content",
+        "engagement_score": 10.0,
+        "relevance_score": 0.7,
+    }
+
+    store.store_findings(run_id, topic["id"], [finding])
+    store.store_findings(
+        run_id,
+        topic["id"],
+        [{**finding, "source_title": "Reddit 1 updated", "engagement_score": 15.0}],
+    )
+
+    sightings = store.get_sightings_for_run(topic["id"], run_id)
+    assert len(sightings) == 1
+    assert sightings[0]["source_title"] == "Reddit 1 updated"
+    assert sightings[0]["engagement_score"] == 15.0
+
+
 def test_update_validates_allowed_columns(temp_db, sample_report):
     """Test update_run/update_finding accept valid keys and reject invalid keys."""
     topic = store.add_topic("Test Topic")
@@ -503,16 +685,16 @@ def test_get_new_findings_filters_by_date(temp_db, sample_report):
     findings = store.findings_from_report(sample_report)
     store.store_findings(run_id, topic["id"], findings)
     
-    # Get findings since tomorrow (should be empty)
+    # Use UTC because store writes first_seen via SQLite's datetime('now') (UTC).
-    tomorrow = (datetime.now() + timedelta(days=1)).strftime("%Y-%m-%d")
+    # Local-time math here would flake near midnight UTC.
+    tomorrow = (datetime.now(timezone.utc) + timedelta(days=1)).strftime("%Y-%m-%d")
     new_findings = store.get_new_findings(topic["id"], since=tomorrow)
-    
+
     assert len(new_findings) == 0
-    
+
-    # Get findings since yesterday (should have all)
+    yesterday = (datetime.now(timezone.utc) - timedelta(days=1)).strftime("%Y-%m-%d")
-    yesterday = (datetime.now() - timedelta(days=1)).strftime("%Y-%m-%d")
     new_findings = store.get_new_findings(topic["id"], since=yesterday)
-    
+
     assert len(new_findings) == 4
 
 

tests/test_version_consistency.py

@@ -1,4 +1,5 @@
 import re
+import sys
 import unittest
 from pathlib import Path
 
@@ -6,16 +7,31 @@ from pathlib import Path
 ROOT = Path(__file__).resolve().parents[1]
 SKILL_ROOT = ROOT / "skills" / "last30days"
 
+sys.path.insert(0, str(SKILL_ROOT / "scripts"))
+from lib.skill_meta import read_skill_version  # noqa: E402
+
 
 def _skill_version() -> str:
-    text = (SKILL_ROOT / "SKILL.md").read_text(encoding="utf-8")
+    version = read_skill_version(SKILL_ROOT / "SKILL.md")
-    match = re.search(r'^version:\s*"([^"]+)"\s*$', text, re.MULTILINE)
+    if not version:
-    if not match:
         raise AssertionError("SKILL.md version frontmatter not found")
-    return match.group(1)
+    return version
 
 
 class TestVersionConsistency(unittest.TestCase):
+    def test_skill_md_uses_double_quoted_version(self) -> None:
+        # The shared VERSION_RE in skill_meta.py accepts double-quoted,
+        # single-quoted, and unquoted YAML version scalars. This repo's
+        # SKILL.md must use the double-quoted form so the badge string stays
+        # deterministic and contributors don't accidentally introduce a
+        # quoting style that's harder for downstream tooling to parse.
+        text = (SKILL_ROOT / "SKILL.md").read_text(encoding="utf-8")
+        self.assertRegex(
+            text,
+            re.compile(r'^version:\s*"[^"]+"\s*$', re.MULTILINE),
+            msg="SKILL.md frontmatter version must use double-quoted form",
+        )
+
     def test_root_skill_header_matches_frontmatter_version(self) -> None:
         text = (SKILL_ROOT / "SKILL.md").read_text(encoding="utf-8")
         version = _skill_version()

tests/test_watchlist_commands.py

@@ -251,7 +251,38 @@ def test_run_topic_success(mock_subprocess, temp_db):
             "source_weights": {},
         },
         "clusters": [],
-        "ranked_candidates": [],
+        "ranked_candidates": [
+            {
+                "candidate_id": "c-r1",
+                "item_id": "R1",
+                "source": "reddit",
+                "title": "Test",
+                "url": "https://reddit.com/1",
+                "snippet": "Snippet",
+                "subquery_labels": ["primary"],
+                "native_ranks": {"reddit": 1},
+                "local_relevance": 0.8,
+                "freshness": 100,
+                "engagement": 50.0,
+                "source_quality": 0.8,
+                "rrf_score": 1.0,
+                "final_score": 0.8,
+                "explanation": "Snippet",
+                "source_items": [
+                    {
+                        "item_id": "R1",
+                        "source": "reddit",
+                        "title": "Test",
+                        "body": "Content",
+                        "url": "https://reddit.com/1",
+                        "author": "user",
+                        "engagement_score": 50.0,
+                        "local_relevance": 0.8,
+                        "snippet": "Snippet",
+                    }
+                ],
+            }
+        ],
         "items_by_source": {
             "reddit": [
                 {
@@ -338,7 +369,38 @@ def test_run_topic_calls_delivery(mock_deliver, mock_subprocess, temp_db):
             "source_weights": {},
         },
         "clusters": [],
-        "ranked_candidates": [],
+        "ranked_candidates": [
+            {
+                "candidate_id": "c-r1",
+                "item_id": "R1",
+                "source": "reddit",
+                "title": "Test",
+                "url": "https://reddit.com/1",
+                "snippet": "Snippet",
+                "subquery_labels": ["primary"],
+                "native_ranks": {"reddit": 1},
+                "local_relevance": 0.8,
+                "freshness": 100,
+                "engagement": 50.0,
+                "source_quality": 0.8,
+                "rrf_score": 1.0,
+                "final_score": 0.8,
+                "explanation": "Snippet",
+                "source_items": [
+                    {
+                        "item_id": "R1",
+                        "source": "reddit",
+                        "title": "Test",
+                        "body": "Content",
+                        "url": "https://reddit.com/1",
+                        "author": "user",
+                        "engagement_score": 50.0,
+                        "local_relevance": 0.8,
+                        "snippet": "Snippet",
+                    }
+                ],
+            }
+        ],
         "items_by_source": {
             "reddit": [
                 {

tests/test_youtube_yt.py

@@ -1,6 +1,7 @@
 """Tests for YouTube transcript highlights and yt-dlp safety flags."""
 
 import json
+import os
 import sys
 import tempfile
 import unittest
@@ -407,5 +408,134 @@ class TestSearchAndTranscribe(unittest.TestCase):
         ft_mock.assert_not_called()
 
 
+class TestYtdlpSSHRouting(unittest.TestCase):
+    """LAST30DAYS_YOUTUBE_SSH_HOST routes yt-dlp invocations through SSH for residential IP."""
+
+    def setUp(self):
+        # Ensure clean env for each test
+        self._saved_env = os.environ.pop("LAST30DAYS_YOUTUBE_SSH_HOST", None)
+
+    def tearDown(self):
+        os.environ.pop("LAST30DAYS_YOUTUBE_SSH_HOST", None)
+        if self._saved_env is not None:
+            os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = self._saved_env
+
+    def test_no_env_var_returns_none(self):
+        """Without the env var set, _ytdlp_ssh_host returns None."""
+        self.assertIsNone(youtube_yt._ytdlp_ssh_host())
+
+    def test_env_var_returns_host(self):
+        """With LAST30DAYS_YOUTUBE_SSH_HOST set, _ytdlp_ssh_host returns it."""
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = "macmini"
+        self.assertEqual(youtube_yt._ytdlp_ssh_host(), "macmini")
+
+    def test_env_var_whitespace_stripped(self):
+        """Whitespace around the host alias is stripped."""
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = "  macmini  "
+        self.assertEqual(youtube_yt._ytdlp_ssh_host(), "macmini")
+
+    def test_empty_env_var_falls_back_to_none(self):
+        """An empty env var is treated as unset."""
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = ""
+        self.assertIsNone(youtube_yt._ytdlp_ssh_host())
+
+    def test_wrap_cmd_passthrough_when_unset(self):
+        """_wrap_ytdlp_cmd returns input unchanged when SSH routing is off."""
+        cmd = ["yt-dlp", "--ignore-config", "ytsearch5:test"]
+        self.assertEqual(youtube_yt._wrap_ytdlp_cmd(cmd), cmd)
+
+    def test_wrap_cmd_prepends_ssh_when_set(self):
+        """_wrap_ytdlp_cmd prepends ssh <host> when SSH routing is on."""
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = "macmini"
+        cmd = ["yt-dlp", "--ignore-config", "ytsearch5:test"]
+        wrapped = youtube_yt._wrap_ytdlp_cmd(cmd)
+        self.assertEqual(wrapped[0], "ssh")
+        self.assertEqual(wrapped[1], "-o")
+        self.assertEqual(wrapped[2], "BatchMode=yes")
+        # `--` terminates SSH option parsing so a host starting with `-`
+        # (e.g. `-oProxyCommand=...`) cannot be reinterpreted as a flag.
+        self.assertEqual(wrapped[3], "--")
+        self.assertEqual(wrapped[4], "macmini")
+        # Final arg is the shell-quoted command string
+        self.assertIn("yt-dlp", wrapped[5])
+        self.assertIn("ytsearch5:test", wrapped[5])
+
+    def test_wrap_cmd_quotes_args_with_spaces(self):
+        """Args containing spaces or special chars are shell-quoted."""
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = "macmini"
+        cmd = ["yt-dlp", "ytsearch5:hello world", "--dump-json"]
+        wrapped = youtube_yt._wrap_ytdlp_cmd(cmd)
+        # shlex.quote wraps the whole arg in single quotes when it contains spaces
+        self.assertIn("'ytsearch5:hello world'", wrapped[5])
+
+    def test_wrap_cmd_uses_option_terminator(self):
+        """`--` is inserted before host as defense-in-depth even for valid hosts."""
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = "macmini"
+        cmd = ["yt-dlp", "--version"]
+        wrapped = youtube_yt._wrap_ytdlp_cmd(cmd)
+        dash_idx = wrapped.index("--")
+        self.assertEqual(wrapped[dash_idx + 1], "macmini")
+
+    def test_host_alias_with_dash_prefix_is_rejected(self):
+        """A host value starting with `-` is rejected by the alias validator.
+
+        Without validation, ssh could parse `-oProxyCommand=...` as a flag
+        instead of a hostname. The `--` terminator in _wrap_ytdlp_cmd is
+        defense-in-depth; this regex on _ytdlp_ssh_host() rejects the value
+        before it ever reaches the ssh command line.
+        """
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = "-oProxyCommand=evil"
+        self.assertIsNone(youtube_yt._ytdlp_ssh_host())
+        # And the wrap function falls back to the local-execution path.
+        cmd = ["yt-dlp", "--version"]
+        self.assertEqual(youtube_yt._wrap_ytdlp_cmd(cmd), cmd)
+
+    def test_host_alias_with_shell_metacharacters_is_rejected(self):
+        """Host values containing spaces, semicolons, $, etc. are rejected."""
+        for bad in ("host;rm -rf /", "host name", "host$IFS", "host`whoami`", "host&cmd"):
+            os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = bad
+            self.assertIsNone(
+                youtube_yt._ytdlp_ssh_host(),
+                msg=f"validator should reject {bad!r}",
+            )
+
+    def test_host_alias_validator_accepts_realistic_aliases(self):
+        """Valid SSH config aliases are accepted: bare names, FQDNs, IPs."""
+        for good in ("macmini", "home-server", "pi5.local", "192.168.1.10", "homelab_box"):
+            os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = good
+            self.assertEqual(youtube_yt._ytdlp_ssh_host(), good)
+
+    def test_is_ytdlp_installed_short_circuits_with_ssh(self):
+        """is_ytdlp_installed returns True without local check when SSH routing is on."""
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = "macmini"
+        with mock.patch("lib.youtube_yt.shutil.which", return_value=None) as which_mock:
+            self.assertTrue(youtube_yt.is_ytdlp_installed())
+            which_mock.assert_not_called()
+
+    def test_is_ytdlp_installed_falls_through_without_ssh(self):
+        """is_ytdlp_installed checks PATH normally when SSH routing is off."""
+        with mock.patch("lib.youtube_yt.shutil.which", return_value="/usr/bin/yt-dlp"):
+            self.assertTrue(youtube_yt.is_ytdlp_installed())
+        with mock.patch("lib.youtube_yt.shutil.which", return_value=None):
+            self.assertFalse(youtube_yt.is_ytdlp_installed())
+
+    def test_search_call_routes_through_ssh(self):
+        """search_youtube wraps the yt-dlp invocation when SSH routing is on."""
+        os.environ["LAST30DAYS_YOUTUBE_SSH_HOST"] = "macmini"
+        from lib.subproc import SubprocResult
+        fake_result = SubprocResult(returncode=0, stdout="", stderr="")
+        with mock.patch.object(youtube_yt.subproc, "run_with_timeout",
+                               return_value=fake_result) as run_mock:
+            youtube_yt.search_youtube("test", "2026-02-01", "2026-03-01")
+        cmd = run_mock.call_args.args[0]
+        self.assertEqual(cmd[0], "ssh")
+        self.assertEqual(cmd[3], "--")
+        self.assertEqual(cmd[4], "macmini")
+        # The shell-quoted yt-dlp invocation lives at index 5
+        self.assertIn("yt-dlp", cmd[5])
+        self.assertIn("--ignore-config", cmd[5])
+        self.assertIn("--no-cookies-from-browser", cmd[5])
+
+
 if __name__ == "__main__":
     unittest.main()

uv.lock

@@ -106,7 +106,7 @@ wheels = [
 
 [[package]]
 name = "last30days-skill"
-version = "3.2.3"
+version = "3.2.4"
 source = { virtual = "." }
 
 [package.dev-dependencies]
@@ -119,7 +119,7 @@ dev = [
 
 [package.metadata.requires-dev]
 dev = [
-    { name = "pytest", specifier = ">=9,<10" },
+    { name = "pytest", specifier = ">=9.0.3,<10" },
     { name = "pytest-cov", specifier = ">=7,<8" },
 ]
 
@@ -152,7 +152,7 @@ wheels = [
 
 [[package]]
 name = "pytest"
-version = "9.0.2"
+version = "9.0.3"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "colorama", marker = "sys_platform == 'win32'" },
@@ -161,9 +161,9 @@ dependencies = [
     { name = "pluggy" },
     { name = "pygments" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/d1/db/7ef3487e0fb0049ddb5ce41d3a49c235bf9ad299b6a25d5780a89f19230f/pytest-9.0.2.tar.gz", hash = "sha256:75186651a92bd89611d1d9fc20f0b4345fd827c41ccd5c299a868a05d70edf11", size = 1568901, upload-time = "2025-12-06T21:30:51.014Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/7d/0d/549bd94f1a0a402dc8cf64563a117c0f3765662e2e668477624baeec44d5/pytest-9.0.3.tar.gz", hash = "sha256:b86ada508af81d19edeb213c681b1d48246c1a91d304c6c81a427674c17eb91c", size = 1572165, upload-time = "2026-04-07T17:16:18.027Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/3b/ab/b3226f0bd7cdcf710fbede2b3548584366da3b19b5021e74f5bde2a8fa3f/pytest-9.0.2-py3-none-any.whl", hash = "sha256:711ffd45bf766d5264d487b917733b453d917afd2b0ad65223959f59089f875b", size = 374801, upload-time = "2025-12-06T21:30:49.154Z" },
+    { url = "https://files.pythonhosted.org/packages/d4/24/a372aaf5c9b7208e7112038812994107bc65a84cd00e0354a88c2c77a617/pytest-9.0.3-py3-none-any.whl", hash = "sha256:2c5efc453d45394fdd706ade797c0a81091eccd1d6e4bccfcd476e2b8e0ab5d9", size = 375249, upload-time = "2026-04-07T17:16:16.13Z" },
 ]
 
 [[package]]