Starr-del secret-preflight
Part of shipsafe — offline deploy-safety skills. Every script is stdlib-only Python 3.8+; nothing leaves the machine.
python3 scripts/scan_secrets.py <project_root> [--json]
Two detection layers:
- High-precision patterns for real key formats: AWS, GitHub, Stripe (live vs test severity), OpenAI, Anthropic, Slack, Google, Supabase, private-key blocks, hardcoded JWTs, Postgres URLs with embedded passwords. JWTs are decoded offline: a payload with
"role":"service_role"is escalated to a dedicated Supabase SERVICE_ROLE finding — that key bypasses Row Level Security entirely, so anywhere near client code it means the whole database is public. This single mistake is behind the largest documented vibe-coding breaches. - Entropy analysis on assignments to secret-named variables (SECRET/TOKEN/API_KEY/...), with placeholder filtering to avoid false positives on 'your-api-key-here'.
Also verifies .gitignore actually covers .env files. Local .env contents are intentionally NOT flagged — that's where secrets belong; committed code is what's scanned.
Critical follow-ups when something is found:
- If the key was EVER committed, removing the line is not enough — it lives in git history. The key must be ROTATED. Say this explicitly.
- A finding can be suppressed for a known-safe line by adding the comment
secret-preflight: allow.
Exit codes: 0 clean, 1 findings.
All paths below are relative to this skill's directory (secret-preflight/).