scrutinizelisted
Install: claude install-skill pitimon/8-habit-ai-dev
# Scrutinize (มองจากภายนอก แล้วไล่ทางจริง)
**Habit**: H5 — Seek First to Understand + H8 — Find Your Voice | **Anti-pattern**: Rubber-stamp "LGTM" reviews that only check what's already in the diff
Stand outside the change and ask whether it should exist at all, then verify it actually does what it claims end-to-end. Different from `/review-ai`: that skill checks security/quality/perf on the diff. This skill questions scope, traces the call graph, and surfaces whether a simpler alternative was missed.
## Operating Stance
- **Outsider.** Forget who wrote it and why they think it's right. Read the artifact cold.
- **End-to-end, not diff-local.** The diff is the entry point, not the scope. Follow the call graph through real code paths.
- **Actionable, concise, with rationale.** Every finding states _what to change_, _why_, and _what evidence_ led you there. No filler, no restating the diff back.
## When to Use
- Before committing to an implementation approach (after `/design` or `/build-brief`, before code is written).
- Before merging a PR — as a second pass alongside `/review-ai`.
- When a plan or design doc feels "fine" but you want an independent check.
- When the user says "scrutinize", "sanity-check", "second opinion", "outsider review", "audit this approach".
## When to Skip
- Auto-generated formatting changes (lint --fix, prettier) — nothing to question.
- Dependency version bumps with passing CI and no API surface change.
- Reverting a commit to its exact previo