Merge pull request #368 from hnshah/ren/advisory-security-workflow-252

ci: add advisory security workflow
This commit is contained in:
Trevin Chow
2026-05-16 23:40:28 -07:00
committed by GitHub
3 changed files with 126 additions and 0 deletions
+67
View File
@@ -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
+6
View File
@@ -27,6 +27,12 @@ npx skills add . -g -y # one-time: symlink this repo into every detected harne
- 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. - 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`) - 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 ## 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`. 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`.
+53
View File
@@ -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