01
What this skill does
Use this skill for an implementation-quality gate. It reviews the declared scope, enumerates distinct hack risks, protects justified exceptions, and records ownership coverage. Without a named scope it prefers staged changes, then the working tree against HEAD. Review-only requests end at the report; authorized repairs continue after the report is preserved. Invocation is explicit-only: a user must invoke `$hack-review`; matching prompts do not activate it automatically.
02
When to use it
- 01Finding impossible-state fallbacks that hide broken invariants.
- 02Flagging symptom-masking patches that do not address root cause.
- 03Catching duplicate abstractions, hardcoded special cases, and boundary bypasses.
03
How it works
- 01
Set the review scope; if unspecified, prefer staged changes and otherwise state the working-tree versus HEAD assumption.
- 02
Read relevant diffs, requirements, and touched ownership boundaries.
- 03
Identify hack-risk patterns and group them into distinct findings.
- 04
Write a Markdown report with recommendation, findings, intentional exceptions, and coverage ledger.
- 05
Keep the gate aligned with the highest-severity unresolved finding and coverage state.
04
What you get back
- 01A coverage-led Markdown hack-risk report.
- 02A short terminal summary.
- 03A complete index of findings, intentional exceptions, and uncovered boundaries.
05
Important boundaries
- 01Do not sample large scopes silently.
- 02Do not lower the recommendation below the strongest unresolved finding.
- 03Use `regression-review` instead when the main question is user-visible behavior.
06