Drop "skills": ["./"] to fix v2.1.105 load regression

The harness rejects "./" in the skills array with
  Path escapes plugin directory: ./ (skills)
even though 2.1.94's changelog explicitly sanctioned this pattern.
This prevents the plugin from loading on current releases.

"./" also created a latent duplicate name: it registered the root
SKILL.md as "last30days", while auto-discovery of
skills/last30days-nux/SKILL.md (a symlink to ../../SKILL.md) registered
the same content under the same name. The duplicate only surfaced once
the registration succeeded.

Dropping "./" makes skills/ auto-discovery the single source for the
harness: skills/last30days-nux/ registers as "last30days", and
skills/last30days/ registers as "last30days-v3-spec" via its frontmatter
name. The root SKILL.md continues to serve the other harnesses that
rely on it (.agents/, .hermes-plugin/, .codex-plugin/, gemini-extension).

.agents/skills/last30days/SKILL.md

@@ -59,7 +59,7 @@ metadata:
       - clawhub
 ---
 
-# last30days v2.9.5: Research Any Topic from the Last 30 Days
+# last30days v3.0.0: Research Any Topic from the Last 30 Days
 
 > **Permissions overview:** Reads public web/platform data and optionally saves research briefings 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.
 

.claude-plugin/plugin.json

@@ -11,6 +11,5 @@
   "repository": "https://github.com/mvanhorn/last30days-skill",
   "license": "MIT",
   "keywords": ["research", "reddit", "twitter", "youtube", "tiktok", "instagram", "trends", "prompts", "polymarket", "github", "perplexity", "threads", "pinterest", "eli5", "hacker-news"],
-  "skills": ["./"],
   "hooks": {}
 }

.gitignore

@@ -15,3 +15,7 @@ variants/open/references/research.md
 __pycache__/
 *.pyc
 mise.toml
+.memsearch/
+.venv/
+.coverage
+htmlcov/

.hermes-plugin/SKILL.md

@@ -0,0 +1,269 @@
+---
+name: last30days
+version: "3.0.0"
+description: "Multi-query social search with intelligent planning. Research any topic across Reddit, X, YouTube, TikTok, Instagram, Hacker News, Polymarket, and the web."
+argument-hint: 'last30days AI video tools, last30days best noise cancelling headphones'
+allowed-tools: Bash, Read, Write, AskUserQuestion, WebSearch
+homepage: https://github.com/mvanhorn/last30days-skill
+repository: https://github.com/mvanhorn/last30days-skill
+author: mvanhorn
+license: MIT
+user-invocable: true
+metadata:
+  hermes:
+    emoji: "📰"
+    tags:
+      - research
+      - deep-research
+      - reddit
+      - x
+      - twitter
+      - youtube
+      - tiktok
+      - instagram
+      - hackernews
+      - polymarket
+      - trends
+      - recency
+      - news
+      - citations
+      - multi-source
+      - social-media
+      - analysis
+      - web-search
+    requires:
+      env:
+        - SCRAPECREATORS_API_KEY
+      optionalEnv:
+        - OPENAI_API_KEY
+        - XAI_API_KEY
+        - OPENROUTER_API_KEY
+        - PARALLEL_API_KEY
+        - BRAVE_API_KEY
+        - APIFY_API_TOKEN
+        - AUTH_TOKEN
+        - CT0
+        - BSKY_HANDLE
+        - BSKY_APP_PASSWORD
+        - TRUTHSOCIAL_TOKEN
+      bins:
+        - node
+        - python3
+    primaryEnv: SCRAPECREATORS_API_KEY
+    files:
+      - "scripts/*"
+    homepage: https://github.com/mvanhorn/last30days-skill
+---
+
+# last30days v3.0.0: Research Any Topic from the Last 30 Days
+
+> **Permissions overview:** Reads public web/platform data and optionally saves research briefings 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.
+
+Research ANY topic across Reddit, X, YouTube, and other sources. Surface what people are actually discussing, recommending, betting on, and debating right now.
+
+## Runtime Preflight
+
+Before running any `last30days.py` command in this skill, resolve a Python 3.12+ interpreter once and keep it in `LAST30DAYS_PYTHON`:
+
+```bash
+for py in python3.14 python3.13 python3.12 python3; do
+  command -v "$py" >/dev/null 2>&1 || continue
+  "$py" -c 'import sys; raise SystemExit(0 if sys.version_info >= (3, 12) else 1)' || continue
+  LAST30DAYS_PYTHON="$py"
+  break
+done
+
+if [ -z "${LAST30DAYS_PYTHON:-}" ]; then
+  echo "ERROR: last30days v3 requires Python 3.12+. Install python3.12 or python3.13 and rerun." >&2
+  exit 1
+fi
+```
+
+## Step 0: First-Run Setup Wizard
+
+**CRITICAL: ALWAYS execute Step 0 BEFORE Step 1, even if the user provided a topic.** If the user typed `last30days Mercer Island`, you MUST check for FIRST_RUN and present the wizard BEFORE running research. The topic "Mercer Island" is preserved — research runs immediately after the wizard completes. Do NOT skip the wizard because a topic was provided. The wizard takes 10 seconds and only runs once ever.
+
+To detect first run: check if `~/.config/last30days/.env` exists. If it does NOT exist, this is a first run. **Do NOT run any Bash commands or show any command output to detect this — just check the file existence silently.** If the file exists and contains `SETUP_COMPLETE=true`, skip this section **silently** and proceed to Step 1. **Do NOT say "Setup is complete" or any other status message — just move on.** The user doesn't need to be told setup is done every time they run the skill.
+
+**When first run is detected, detect your platform first:**
+
+**If you do NOT have WebSearch capability (raw CLI):** Run the terminal-only setup flow below.
+**If you DO have WebSearch (Hermes):** Run the standard setup flow below.
+
+---
+
+### Terminal-Only / Non-WebSearch Setup Flow
+
+Run environment detection first:
+```bash
+"${LAST30DAYS_PYTHON}" "${SKILL_ROOT}/scripts/last30days.py" setup --terminal
+```
+
+Read the JSON output. It tells you what's already configured. Display a status summary:
+
+```
+👋 Welcome to last30days!
+
+Detected:
+{✅ or ❌} yt-dlp (YouTube search)
+{✅ or ❌} X/Twitter ({method} configured)
+{✅ or ❌} ScrapeCreators (TikTok, Instagram, Reddit backup)
+{✅ or ❌} Web search ({backend} configured)
+```
+
+Then for each missing item, offer setup in priority order:
+
+1. **ScrapeCreators** (if not configured): "ScrapeCreators adds TikTok and Instagram search (plus a Reddit backup if public Reddit gets rate-limited). 10,000 free calls, no credit card. (No referrals, no kickbacks - we don't get a cut.)"
+   - Option A: "ScrapeCreators via GitHub (recommended)" — Check if `gh` CLI was detected in the environment detection output above. If gh IS detected: description should say "Registers directly via GitHub CLI in ~2 seconds - no browser needed". Before running the command, display: "Registering via GitHub CLI..." If gh is NOT detected: description should say "Copies a one-time code to your clipboard and opens GitHub to authorize". Then run `"${LAST30DAYS_PYTHON}" "${SKILL_ROOT}/scripts/last30days.py" setup --github`, parse JSON output. Tries PAT first (if `gh` is installed), falls back to device flow which copies a one-time code to your clipboard and opens your browser. If `status` is `success`, write `SCRAPECREATORS_API_KEY=*** to .env.
+   - Option B: "I have a key" — accept paste, write to .env
+   - Option C: "Skip for now"
+
+2. **X/Twitter** (if not configured): "X search finds tweets and conversations. To unlock X: add FROM_BROWSER=auto (reads browser cookies, free), XAI_API_KEY (no browser access, api.x.ai), or AUTH_TOKEN+CT0 (manual cookies)."
+   - Option A: "I have an xAI API key" (recommended for servers — persistent, no expiry). Write XAI_API_KEY to .env.
+   - Option B: "I have AUTH_TOKEN + CT0 from my browser" — accept both, write to .env
+   - Option C: "Skip for now"
+
+3. **YouTube** (if yt-dlp not found): "YouTube search needs yt-dlp. Run: `pip install yt-dlp`"
+
+4. **Web search** (if no Brave/Exa/Serper key): "A web search key enables smarter results. Brave Search is free for 2,000 queries/month at brave.com/search/api"
+
+After setup, write `SETUP_COMPLETE=true` to .env and proceed to research.
+
+**Skip to "END OF FIRST-RUN WIZARD" below after completing the terminal-only flow.**
+
+---
+
+### Hermes Setup Flow (Standard)
+
+**You MUST follow these steps IN ORDER. Do NOT skip ahead to the topic picker or research. The sequence is: (1) welcome text -> (2) setup modal -> (3) run setup if chosen -> (4) optional ScrapeCreators modal -> (5) topic picker. You MUST start at step 1.**
+
+**Step 1: Display the following welcome text ONCE as a normal message (not blockquoted). Then IMMEDIATELY call AskUserQuestion - do NOT repeat any of the welcome text inside the AskUserQuestion call.**
+
+Welcome to last30days!
+
+I research any topic across Reddit, X, YouTube, and other sources - synthesizing what people are actually saying right now.
+
+Auto setup gives you 5 core sources for free in 30 seconds:
+- X/Twitter - reads your x.com browser cookies to authenticate (not saved to disk). Chrome on macOS will prompt for Keychain access.
+- Reddit with comments - public JSON, no API key needed
+- YouTube search + transcripts - installs yt-dlp (open source, 190K+ GitHub stars)
+- Hacker News + Polymarket + GitHub (if `gh` CLI installed) - always on, zero config
+
+Want TikTok and Instagram too? ScrapeCreators adds those (10,000 free calls, scrapecreators.com). No kickbacks, no affiliation.
+
+**Then call AskUserQuestion with ONLY this question and these options - no additional text:**
+
+Question: "How would you like to set up?"
+Options:
+- "Auto setup (~30 seconds) - scans browser cookies for X + installs yt-dlp for YouTube"
+- "Manual setup - show me what to configure"
+- "Skip for now - Reddit (with comments), HN, Polymarket, GitHub (if gh installed), Web"
+
+**If the user picks 1 (Auto setup):**
+
+**Before running the setup command, get cookie consent:**
+
+Check if `BROWSER_CONSENT=true` already exists in `~/.config/last30days/.env`. If it does, skip the consent prompt and run setup directly.
+
+If `BROWSER_CONSENT=true` is NOT present, **call AskUserQuestion:**
+Question: "Auto setup will scan your browser for x.com cookies to authenticate X search. Cookies are read live, not saved to disk. Chrome on macOS will prompt for Keychain access. OK to proceed?"
+Options:
+- "Yes, scan my cookies for X" - Run setup as normal. Append `BROWSER_CONSENT=true` to .env after setup completes.
+- "Skip X, just set up YouTube" - Run setup with YouTube only (install yt-dlp). Do not scan cookies.
+- "I have an xAI API key instead" - Ask them to paste it, write XAI_API_KEY to .env. Then install yt-dlp.
+
+Run the setup subcommand:
+```bash
+cd {SKILL_DIR} && "${LAST30DAYS_PYTHON}" scripts/last30days.py setup
+```
+Show the user the results (what cookies were found, whether yt-dlp was installed).
+
+**Then show the optional ScrapeCreators offer (plain text, then modal):**
+
+Want TikTok and Instagram too? ScrapeCreators adds those platforms - 10,000 free calls, no credit card. It also serves as a Reddit backup if public Reddit ever gets rate-limited.
+
+**Before showing the ScrapeCreators modal, check for `gh` CLI:** Run `which gh` via Bash silently. Store the result as gh_available (true if found, false if not).
+
+**Call AskUserQuestion:**
+Question: "Want to add TikTok, Instagram, and Reddit backup via ScrapeCreators? (We don't get a cut.)"
+Options:
+- "ScrapeCreators via GitHub (fastest, recommended)" - If gh_available: description should say "Registers directly via GitHub CLI in ~2 seconds - no browser needed". If NOT gh_available: description should say "Copies a one-time code to your clipboard and opens GitHub to authorize". After the user selects this option: If gh_available, display "Registering via GitHub CLI..." before running the command. If NOT gh_available, display "I'll copy a one-time code to your clipboard and open GitHub. When GitHub asks for a device code, just paste (Cmd+V on Mac, Ctrl+V on Windows/Linux)." Then run `cd {SKILL_DIR} && "${LAST30DAYS_PYTHON}" scripts/last30days.py setup --github` via Bash with a 5-minute timeout. This tries PAT auth first (if `gh` CLI is installed, zero browser needed), then falls back to GitHub device flow which copies a one-time code to your clipboard and opens GitHub in your browser. Parse the JSON stdout. If `status` is `success`, write `SCRAPECREATORS_API_KEY=*** to `~/.config/last30days/.env`. If `method` is `pat`, show: "You're in! Registered via GitHub CLI - zero browser needed. 10,000 free calls. TikTok, Instagram, and Reddit backup are now active." If `method` is `device` and `clipboard_ok` is true, show: "You're in! (The authorization code was copied to your clipboard automatically.) 10,000 free calls. TikTok, Instagram, and Reddit backup are now active." If `method` is `device` and `clipboard_ok` is false, show: "You're in! 10,000 free calls. TikTok, Instagram, and Reddit backup are now active." If `status` is `timeout` or `error`, show: "GitHub auth didn't complete. No worries - you can sign up at scrapecreators.com instead or try again later." Then offer the web signup option.
+- "Open scrapecreators.com (Google sign-in)" - run `open https://scrapecreators.com` via Bash to open in the user's browser. Then ask them to paste the API key they get. When they paste it, write SCRAPECREATORS_API_KEY=*** to ~/.config/last30days/.env
+- "I have a key" - accept the key, write to .env
+- "Skip for now" - proceed without ScrapeCreators
+
+**After SC key is saved (not if skipped), show the TikTok/Instagram opt-in:**
+
+**Call AskUserQuestion:**
+Question: "Enable TikTok and Instagram search?"
+Options:
+- "Yes, enable TikTok + Instagram" - Write `TIKTOK_ENABLED=true` and `INSTAGRAM_ENABLED=true` to .env. Then show: "TikTok and Instagram are now enabled. You can disable them later by editing ~/.config/last30days/.env."
+- "No, skip for now" - proceed without enabling
+
+**After setup completes, write `SETUP_COMPLETE=true` to .env.**
+
+---
+
+## END OF FIRST-RUN WIZARD
+
+Proceed to Step 1.
+
+---
+
+## Step 1: Parse Topic
+
+The user invoked: `last30days {QUERY}`
+
+Extract the topic. If the query is empty or ambiguous, ask for clarification.
+
+## Step 2: Execute Research
+
+Run the research engine:
+
+```bash
+cd {SKILL_DIR} && "${LAST30DAYS_PYTHON}" scripts/last30days.py "{TOPIC}" --emit=compact --lookback-days=30
+```
+
+Optional flags based on user request:
+- `--search=reddit,youtube,hackernews` - Specific sources only
+- `--days=7` - Shorter time range
+- `--deep` - Higher recall mode
+- `--save` - Save to ~/Documents/Last30Days/
+
+## Step 3: Display Results
+
+Show the research output to the user. The compact output includes:
+- Executive summary
+- Ranked evidence clusters with scores
+- Source statistics (upvotes, views, engagement)
+- Citations with URLs
+- Confidence levels and uncertainty notes
+
+## Security & Permissions
+
+**What this skill does:**
+- Sends search queries to ScrapeCreators API (`api.scrapecreators.com`) for TikTok and Instagram search, and as a Reddit backup when public Reddit is unavailable (requires SCRAPECREATORS_API_KEY)
+- Sends search queries to OpenAI's Responses API (`api.openai.com`) for Reddit discovery (fallback if no SCRAPECREATORS_API_KEY)
+- Sends search queries to Twitter's GraphQL API (via optional user-provided AUTH_TOKEN/CT0 env vars — no browser session access) or xAI's API (`api.x.ai`) for X search
+- 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)
+- 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)
+- Saves research briefings as .md files to ~/Documents/Last30Days/
+
+**What this skill does NOT do:**
+- Does not post, like, or modify content on any platform
+- Does not access your Reddit, X, or YouTube accounts
+- Does not share API keys between providers (OpenAI key only goes to api.openai.com, etc.)
+- 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.
+- 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)
+
+Review scripts before first use to verify behavior.

CHANGELOG.md

@@ -189,7 +189,6 @@ Three headline features: watchlists for always-on bots, YouTube transcripts as a
 
 ### Credits
 
-- @steipete -- Bird CLI (vendored X search) and yt-dlp/summarize inspiration for YouTube transcripts
 - @galligan -- Marketplace plugin inspiration
 - @hutchins -- Pushed for YouTube feature
 

HERMES_SETUP.md

@@ -0,0 +1,121 @@
+# Hermes Setup Guide for last30days
+
+This guide covers installing last30days on Hermes AI Agent.
+
+## Prerequisites
+
+1. **Hermes installed** - See https://github.com/mercurial-tf/hermes
+2. **Python 3.12+** - `brew install python@3.12` or similar
+3. **yt-dlp** (optional, for YouTube) - `brew install yt-dlp`
+
+## Installation
+
+### Option 1: Via sync.sh (Recommended)
+
+```bash
+# Clone the repo
+git clone https://github.com/mvanhorn/last30days-skill.git
+cd last30days-skill
+
+# Run the sync script
+bash scripts/sync.sh
+```
+
+This will auto-detect Hermes and deploy to `~/.hermes/skills/research/last30days/`
+
+### Option 2: Manual Copy
+
+```bash
+# Create directory
+mkdir -p ~/.hermes/skills/research/last30days
+
+# Copy files
+cp -r scripts ~/.hermes/skills/research/last30days/
+cp .hermes-plugin/SKILL.md ~/.hermes/skills/research/last30days/
+```
+
+## Usage
+
+In Hermes, invoke with:
+
+```
+last30days "your research topic"
+```
+
+Or with options:
+```
+last30days "best mechanical keyboards 2025" --search=reddit,youtube
+last30days "AI news" --days=7 --deep
+```
+
+## First Run Setup
+
+On first run, the skill will guide you through setup:
+
+1. **Auto setup** (~30 seconds)
+   - Scans browser cookies for X/Twitter
+   - Checks/installs yt-dlp for YouTube
+   - Configures free sources (Reddit, HN, Polymarket)
+
+2. **Optional: ScrapeCreators**
+   - Adds TikTok, Instagram, Reddit backup
+   - 10,000 free API calls
+   - Sign up at scrapecreators.com
+
+3. **Optional: API Keys**
+   - XAI_API_KEY for X/Twitter (alternative to browser cookies)
+   - BRAVE_API_KEY for web search
+
+## Available Sources
+
+### Free (No API Key)
+- **Reddit** - Public discussions and comments
+- **Hacker News** - Tech discussions via Algolia
+- **Polymarket** - Prediction markets
+- **YouTube** - Search and transcripts (requires yt-dlp)
+
+### Requires API Key
+- **X/Twitter** - xAI API key or browser cookies
+- **TikTok** - ScrapeCreators API
+- **Instagram** - ScrapeCreators API
+- **Web Search** - Brave Search API
+
+## Troubleshooting
+
+### Python not found
+```bash
+# Find Python 3.12+
+which python3.12 python3.13 python3.14
+
+# If not installed
+brew install python@3.12
+```
+
+### yt-dlp not found
+```bash
+brew install yt-dlp
+# or
+pip install yt-dlp
+```
+
+### Check what's configured
+```bash
+cd ~/.hermes/skills/research/last30days
+python3.12 scripts/last30days.py --diagnose
+```
+
+## Updating
+
+To update to the latest version:
+
+```bash
+cd last30days-skill
+git pull
+bash scripts/sync.sh
+```
+
+## Support
+
+- Original repo: https://github.com/mvanhorn/last30days-skill
+- Hermes: https://github.com/mercurial-tf/hermes
+- Issues: Please report in the original repo

README.md

@@ -24,6 +24,12 @@ OpenClaw:
 clawhub install last30days-official
 ```
 
+Hermes:
+```
+# The skill auto-deploys when you run sync.sh
+# Or manually copy to ~/.hermes/skills/research/last30days/
+```
+
 Zero config. Reddit, HN, Polymarket, and GitHub work immediately. Run it once and the setup wizard unlocks X, YouTube, TikTok, and more in 30 seconds.
 
 ---
@@ -152,6 +158,25 @@ claude plugin update last30days@last30days-skill
 clawhub install last30days-official
 ```
 
+### Gemini CLI
+
+Gemini CLI supports installing extensions from GitHub repositories, but as of Gemini CLI v0.9.0 there is an upstream installer bug that can fail with:
+
+`Configuration file not found at /tmp/gemini-extensionXXXXXX/gemini-extension.json`
+
+even when `gemini-extension.json` exists at the repo root.
+
+Upstream bug:
+- https://github.com/google-gemini/gemini-cli/issues/11452
+
+Workarounds:
+1) Clone locally, then install from the local path
+   ```bash
+   git clone https://github.com/mvanhorn/last30days-skill
+   gemini extensions install ./last30days-skill
+   ```
+2) If GitHub install fails, use the OpenClaw or Claude Code install paths above.
+
 ### Manual
 ```bash
 git clone https://github.com/mvanhorn/last30days-skill.git ~/.claude/skills/last30days

SKILL.md

@@ -59,7 +59,7 @@ metadata:
       - clawhub
 ---
 
-# last30days v2.9.5: Research Any Topic from the Last 30 Days
+# last30days v3.0.0: Research Any Topic from the Last 30 Days
 
 > **Permissions overview:** Reads public web/platform data and optionally saves research briefings 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.
 

fixtures/eval_topics.json

@@ -0,0 +1,42 @@
+[
+  {
+    "topic": "OpenClaw vs NanoClaw vs ZeroClaw",
+    "query_type": "comparison",
+    "rationale": "Multi-entity extraction, 3-way split across AI agent frameworks."
+  },
+  {
+    "topic": "how to set up a GLP-1 supplement routine",
+    "query_type": "how_to",
+    "rationale": "Trending health topic. Tests non-tech how_to."
+  },
+  {
+    "topic": "2026 March Madness",
+    "query_type": "breaking_news",
+    "rationale": "Live sporting event. Tests broad breaking news recall."
+  },
+  {
+    "topic": "best budget noise cancelling headphones 2026",
+    "query_type": "product",
+    "rationale": "Evergreen consumer query. Tests product review aggregation."
+  },
+  {
+    "topic": "thoughts on OpenAI Codex pricing",
+    "query_type": "opinion",
+    "rationale": "Active developer debate. Tests opinion mining."
+  },
+  {
+    "topic": "odds of US recession 2026",
+    "query_type": "prediction",
+    "rationale": "Major macro topic. Tests prediction market + news synthesis."
+  },
+  {
+    "topic": "what is retrieval augmented generation",
+    "query_type": "concept",
+    "rationale": "Widely discussed AI concept. Tests explanation quality."
+  },
+  {
+    "topic": "Google Wiz acquisition price and timeline",
+    "query_type": "factual",
+    "rationale": "Completed event ($32B). Tests factual precision."
+  }
+]

hooks/scripts/check-config.sh

@@ -12,7 +12,11 @@ check_perms() {
   local file="$1"
   if [[ ! -f "$file" ]]; then return; fi
   local perms
-  perms=$(stat -f '%Lp' "$file" 2>/dev/null || stat -c '%a' "$file" 2>/dev/null || echo "")
+  # Try GNU stat first (Linux), fall back to BSD stat (macOS).
+  # On Linux, `stat -f` prints filesystem info (not permissions) and exits 0,
+  # so the previous BSD-first ordering left $perms as multi-line garbage on
+  # every Linux session start and printed a false WARNING.
+  perms=$(stat -c '%a' "$file" 2>/dev/null || stat -f '%Lp' "$file" 2>/dev/null || echo "")
   if [[ -n "$perms" && "$perms" != "600" && "$perms" != "400" ]]; then
     echo "/last30days: WARNING — $file has permissions $perms (should be 600)."
     echo "  Fix: chmod 600 $file"

release-notes.md

@@ -74,40 +74,12 @@ Contributors who shaped the release itself:
 - @Cody-Coyote (#204) reported the marketplace validation bug that needed fixing before v3 could ship cleanly
 - @dannyshmueli pushed for v3 and Codex family support publicly on X
 
-## What's New
+Full Added / Changed / Fixed detail lives in [CHANGELOG.md](CHANGELOG.md) under `[3.0.0]`.
 
-### Added
+## Earlier contributors
 
-- Intelligent pre-research brain resolving X handles, subreddits, TikTok hashtags, and YouTube channels before searching
-- Fun judge and Best Takes section scoring humor, wit, and virality
-- Cross-source cluster merging with entity-based overlap detection
-- Single-pass comparisons for "X vs Y" queries
-- GitHub as a first-class source with person-mode and project-mode
-- Perplexity Sonar Pro via OpenRouter (`INCLUDE_SOURCES=perplexity`)
-- Perplexity Deep Research (`--deep-research` flag)
-- Parallel AI grounding backend (`--web-backend parallel`)
-- OpenRouter as a reasoning provider (auto-detected after Gemini / OpenAI / xAI)
-- Per-author cap (max 3 items per author)
-- Entity disambiguation trusting resolved handles over keyword matches
-- OpenAI Codex CLI integration via `.agents/skills/last30days/SKILL.md` and `.codex-plugin/plugin.json`
-- ELI5 mode
+From the v1 and v2 lineage:
 
-### Changed
-
-- YouTube transcript candidate pool widened 3x to reach talk and review content with captions
-- Reddit comment enrichment sorted by total engagement (upvotes + comments), not just upvotes
-- Polymarket display shows % odds only, dollar volumes removed
-- 852 tests passing
-
-### Fixed
-
-- Marketplace validation: duplicate `name: last30days` collision in `skills/last30days/SKILL.md` that caused strict validators to reject the plugin. Resolved by renaming the internal v3 architecture spec to `last30days-v3-spec` in #214
-- Stale README link to the deleted `skills/last30days-v3/` path from the v3 directory rename. Fixed in #214
-- Codex CLI discovery: added the real `.agents/skills/last30days/SKILL.md` (regular file, not a symlink, since Codex's loader skips symlinked files) and `.codex-plugin/plugin.json` namespace marker in #219
-
-## Credits
-
-- [@steipete](https://github.com/steipete) for Bird CLI (vendored X search) and yt-dlp/summarize inspiration for YouTube transcripts
 - [@galligan](https://github.com/galligan) for marketplace plugin inspiration
 - [@hutchins](https://x.com/hutchins) for pushing the YouTube feature
 

scripts/last30days.py

@@ -103,7 +103,7 @@ def save_output(report: schema.Report, emit: str, save_dir: str, suffix: str = "
         content = emit_output(report, emit)
     else:
         content = render.render_full(report)
-    out_path.write_text(content)
+    out_path.write_text(content, encoding="utf-8")
     return out_path
 
 
@@ -165,7 +165,14 @@ def build_parser() -> argparse.ArgumentParser:
     parser.add_argument("--tiktok-hashtags", help="Comma-separated TikTok hashtags without # (e.g., tella,screenrecording)")
     parser.add_argument("--tiktok-creators", help="Comma-separated TikTok creator handles (e.g., TellaHQ,taborplace)")
     parser.add_argument("--ig-creators", help="Comma-separated Instagram creator handles (e.g., tella.tv,laborstories)")
-    parser.add_argument("--lookback-days", type=int, default=30, help="Number of days to look back for research (default: 30, watchlist uses 90)")
+    parser.add_argument(
+        "--days",
+        "--lookback-days",
+        dest="lookback_days",
+        type=int,
+        default=30,
+        help="Number of days to look back for research (default: 30, watchlist uses 90)",
+    )
     parser.add_argument("--auto-resolve", action="store_true",
                         help="Use web search to discover subreddits/handles before planning (for platforms without WebSearch)")
     parser.add_argument("--github-user", help="GitHub username for person-mode search (e.g., steipete)")

scripts/lib/bird_x.py

@@ -460,7 +460,7 @@ def parse_bird_response(response: Dict[str, Any], query: str = "") -> List[Dict[
             "url": url,
             "author_handle": author_handle.lstrip("@"),
             "date": date,
-            "engagement": engagement,
+            "engagement": engagement if any(v is not None for v in engagement.values()) else None,
             "why_relevant": "",  # Bird doesn't provide relevance explanations
             "relevance": _compute_relevance(query, str(tweet.get("text", ""))) if query else 0.7,
         }

scripts/sync.sh

@@ -11,7 +11,7 @@ COMMON_TARGETS=(
   # but local development needs the cache kept in sync with the repo.
   # Do NOT add ~/.claude/skills/last30days - it creates a duplicate
   # /last30days-3 in the slash command menu alongside the plugin version.
-  "$HOME/.claude/plugins/cache/last30days-skill-private/last30days-3/3.0.0-alpha"
+  "$HOME/.claude/plugins/cache/last30days-skill-private/last30days-3/3.0.0"
   "$HOME/.claude/plugins/cache/last30days-skill-private/last30days-3-nogem/3.0.0-nogem"
   "$HOME/.agents/skills/last30days"
   "$HOME/.codex/skills/last30days"
@@ -24,7 +24,7 @@ sync_target() {
 
   echo ""
   echo "--- Syncing to $target ---"
-  mkdir -p "$target/scripts/lib" "$target/variants/open/references"
+  mkdir -p "$target/scripts/lib"
 
   cp "$skill_md" "$target/SKILL.md"
 
@@ -35,7 +35,13 @@ sync_target() {
     "$SRC/scripts/store.py" \
     "$target/scripts/"
   rsync -a "$SRC/scripts/lib/"*.py "$target/scripts/lib/"
-  rsync -a "$SRC/variants/open/" "$target/variants/open/"
+
+  # The OpenClaw variant lives in the private repo only. Skip cleanly when
+  # running this script from the public repo where variants/open does not exist.
+  if [ -d "$SRC/variants/open" ]; then
+    mkdir -p "$target/variants/open/references"
+    rsync -a "$SRC/variants/open/" "$target/variants/open/"
+  fi
 
   if [ -d "$SRC/scripts/lib/vendor" ]; then
     rsync -a "$SRC/scripts/lib/vendor" "$target/scripts/lib/"
@@ -63,7 +69,60 @@ for t in "${COMMON_TARGETS[@]}"; do
   sync_target "$t" "$SRC/SKILL.md"
 done
 
-sync_target "$OPENCLAW_TARGET" "$SRC/variants/open/SKILL.md"
+# Hermes sync: deploy to Hermes skills directory if it exists
+HERMES_TARGET="$HOME/.hermes/skills/research/last30days"
+if [ -d "$HOME/.hermes/skills/research" ]; then
+  echo ""
+  echo "--- Syncing to Hermes ---"
+  mkdir -p "$HERMES_TARGET/scripts/lib"
+  
+  # Use Hermes-specific SKILL.md if available, fallback to main
+  if [ -f "$SRC/.hermes-plugin/SKILL.md" ]; then
+    cp "$SRC/.hermes-plugin/SKILL.md" "$HERMES_TARGET/SKILL.md"
+  else
+    cp "$SRC/SKILL.md" "$HERMES_TARGET/SKILL.md"
+  fi
+  
+  rsync -a \
+    "$SRC/scripts/last30days.py" \
+    "$SRC/scripts/watchlist.py" \
+    "$SRC/scripts/briefing.py" \
+    "$SRC/scripts/store.py" \
+    "$HERMES_TARGET/scripts/"
+  rsync -a "$SRC/scripts/lib/"*.py "$HERMES_TARGET/scripts/lib/"
+  
+  if [ -d "$SRC/scripts/lib/vendor" ]; then
+    rsync -a "$SRC/scripts/lib/vendor" "$HERMES_TARGET/scripts/lib/"
+  fi
+  
+  if [ -d "$SRC/fixtures" ]; then
+    mkdir -p "$HERMES_TARGET/fixtures"
+    rsync -a "$SRC/fixtures/" "$HERMES_TARGET/fixtures/"
+  fi
+  
+  mod_count=$(ls "$HERMES_TARGET/scripts/lib/"*.py 2>/dev/null | wc -l | tr -d ' ')
+  echo "  Copied $mod_count modules to Hermes"
+  
+  if (
+    cd "$HERMES_TARGET/scripts" &&
+    python3 -c "import briefing, store, watchlist; from lib import youtube_yt, bird_x, render, ui; print('  Import check: OK')"
+  ); then
+    true
+  else
+    echo "  Import check FAILED"
+  fi
+fi
+
+# OpenClaw sync only runs when the private-repo OpenClaw variant is present
+# in the source tree. The public repo does not ship variants/open (the variant
+# is sanitized via strip_for_openclaw.py and published separately from
+# last30days-skill-private).
+if [ -d "$SRC/variants/open" ]; then
+  sync_target "$OPENCLAW_TARGET" "$SRC/variants/open/SKILL.md"
+else
+  echo ""
+  echo "Skipping OpenClaw target (no variants/open in this repo)"
+fi
 
 echo ""
 echo "Sync complete."

tests/test_bird_x.py

@@ -175,7 +175,7 @@ class TestVendoredBirdRuntime(unittest.TestCase):
             }
         ]
         items = parse_bird_response(tweets, "test query")
-        self.assertIsNone(items[0]["engagement"]["likes"])
+        self.assertIsNone(items[0]["engagement"])
 
     def test_fallback_to_second_key(self):
         tweets = [
@@ -203,6 +203,32 @@ class TestVendoredBirdRuntime(unittest.TestCase):
         items = parse_bird_response(tweets, "test query")
         self.assertEqual(0, items[0]["engagement"]["likes"])
 
+    def test_engagement_none_when_all_fields_missing(self):
+        """All-None engagement dict should become None, not propagate."""
+        tweets = [
+            {
+                "id": "1",
+                "text": "test",
+                "permanent_url": "https://x.com/u/status/1",
+            }
+        ]
+        items = parse_bird_response(tweets, "test query")
+        self.assertIsNone(items[0]["engagement"])
+
+    def test_engagement_preserved_when_any_field_present(self):
+        """Engagement dict kept when at least one metric exists."""
+        tweets = [
+            {
+                "id": "1",
+                "text": "test",
+                "permanent_url": "https://x.com/u/status/1",
+                "likeCount": 5,
+            }
+        ]
+        items = parse_bird_response(tweets, "test query")
+        self.assertIsNotNone(items[0]["engagement"])
+        self.assertEqual(5, items[0]["engagement"]["likes"])
+
 
 if __name__ == "__main__":
     unittest.main()

tests/test_cli_v3.py

@@ -77,6 +77,13 @@ class CliV3Tests(unittest.TestCase):
         with self.assertRaises(SystemExit):
             cli.parse_search_flag(" , ")
 
+    def test_build_parser_accepts_days_alias_and_preserves_topic_tokens(self):
+        parser = cli.build_parser()
+        args, extra = parser.parse_known_args(["--days", "7", "biosecurity", "ai", "agents"])
+        self.assertEqual(7, args.lookback_days)
+        self.assertEqual(["biosecurity", "ai", "agents"], args.topic)
+        self.assertEqual([], extra)
+
     def test_ensure_supported_python_rejects_old_interpreter_with_actionable_error(self):
         stderr = io.StringIO()
         with redirect_stderr(stderr):
@@ -128,6 +135,14 @@ class CliV3Tests(unittest.TestCase):
             payload = json.loads(path.read_text())
             self.assertEqual("OpenClaw vs NanoClaw", payload["topic"])
 
+    def test_save_output_writes_utf8_encoded_markdown(self):
+        report = self.make_report()
+        with tempfile.TemporaryDirectory() as tmp:
+            with mock.patch("pathlib.Path.write_text", autospec=True, return_value=1) as write_text:
+                cli.save_output(report, "md", tmp)
+        _, kwargs = write_text.call_args
+        self.assertEqual("utf-8", kwargs.get("encoding"))
+
     def test_persist_report_updates_run_status_on_success_and_failure(self):
         report = self.make_report()
 

tests/test_version_consistency.py

@@ -0,0 +1,30 @@
+import re
+import unittest
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+
+
+def _skill_version() -> str:
+    text = (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_root_skill_header_matches_frontmatter_version(self) -> None:
+        text = (ROOT / "SKILL.md").read_text(encoding="utf-8")
+        version = _skill_version()
+        self.assertIn(f"# last30days v{version}:", text)
+
+    def test_sync_cache_path_uses_skill_version(self) -> None:
+        sync_text = (ROOT / "scripts" / "sync.sh").read_text(encoding="utf-8")
+        version = _skill_version()
+        self.assertIn(f'last30days-3/{version}"', sync_text)
+
+
+if __name__ == "__main__":
+    unittest.main()