prodcheck

A Claude Code skill that audits your working code for production readiness across 8 dimensions before you ship, demo, or hand off.

What it checks

1. Secret hygiene — hardcoded API keys, .env discipline
2. Error handling — unhandled async, external API call failures
3. Logging & observability — debug logs in hot paths, PII leakage
4. Input validation & injection safety — SQL, prompt, path, shell
5. Configuration — externalized vs hardcoded
6. Code structure — long functions, dead code, magic numbers, duplication
7. Test coverage signal — does it look like anyone thought about testing
8. PR/handoff readiness — self-documenting, clear scope, no leftover TODOs

For each: PASS / WARN / FAIL with line callouts and concrete before/after fixes.

Install

git clone https://github.com/matthewkim323/prodcheck ~/.claude/skills/prodcheck

Use

In any Claude Code session:

prodcheck                    # audit git diff HEAD
prodcheck --staged           # only staged changes (pre-commit)
prodcheck path/to/file.ts    # audit specific file/dir
prodcheck --strict           # add HIPAA/SOC2 checks (PII in logs = FAIL)

Or just say "is this prod ready" / "production check" / "can i ship this".

Output

PRODCHECK REPORT
================
Diff: <what was analyzed>
Checks run: 8

RESULTS:
✅ Secret hygiene — PASS
⚠️  Logging — WARN: unstructured logs in connector.py:34
❌ Input validation — FAIL: user input interpolated raw into LLM prompt at agent.py:112

VERDICT: NOT SHIP-READY (1 FAIL, 1 WARN)

Fix before merging:
[CRITICAL] agent.py:112 — wrap user input in delimiters
  Before: prompt = f"Answer this: {user_query}"
  After:  prompt = f"Answer this: <user_input>{user_query}</user_input>"

License

MIT.