verification-looplisted

Use when verifying shipped work on prod, standing up the autonomous verifier, running the verification loop, confirming a spec is actually done end-to-end, or checking whether shipped criteria are hollow. Trigger phrases include 'verify shipped work', 'stand up the verifier', 'run the verification loop', 'is this actually done on prod', 'check for hollow specs', 'autonomous post-ship review'.
fredhead88/do-it · ★ 0 · AI & Automation · score 72

Install: claude install-skill fredhead88/do-it

# Verification Loop ## Overview **Observe the running product. Stay blind to how it was built.** The verification loop is a standing autonomous reviewer that drives shipped work from "orc says done" to "verified green on prod". It observes the deployed product via a headless browser, assigns typed evidence to each acceptance criterion, judges cross-vendor (Codex primary, Claude fallback), and loops to convergence — filing correctives for hollow work, escalating taste/blockers, and never touching the build. ## Three Core Invariants 1. **Blind-but-watching.** The verifier never sees the build, the diff, or the builder's reasoning. The judge receives only the typed evidence artifact — never the worker's explanation. 2. **Evidence-type-locked-to-criterion-type.** A UI criterion requires a DOM/screenshot observation. A grep is auto-fail for a UI criterion. No criterion closes without observed, type-matched evidence. 3. **Verifier owns the verdict; the builder cannot overwrite it.** Verdicts live in `~/.claude/ledger/verified/<spec_id>.yml` — a separate namespace the builder's `set`/`register` commands never touch. ## The Two-Body Warning The loop converges **only if orc is running and consuming correctives**. If filed items sit unconsumed for N ticks, escalate to `NEEDS-HUMAN.jsonl` — do not file forever into the void. ## The 8-Step Tick Run with: `node /path/to/verification-loop/tick.mjs --config <project> [--spec NNN-slug] [--dry-run] [--force]` 1. **