thkt/polish
attic/skills/polish/SKILL.md
Codex review + cleanup. Findings are challenged by critic-audit, not aggregated as facts, and fixes are applied directly. Do NOT use for internal multi-reviewer deep audits or findings reports (use /audit).
$ install --global
skillsauthnpx skillsauth add thkt/dotclaude polishInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Jul 13, 2026, 2:30 AM71.3s1 file scanned
SKILL.md
- name:
- polish
- description:
- Codex review + cleanup. Findings are challenged by critic-audit, not aggregated as facts, and fixes are applied directly. Do NOT use for internal multi-reviewer deep audits or findings reports (use /audit).
- when_to_use:
- 整理して, きれいにして, コード整理, slop除去, ポリッシュ, テスト整理, テスト監査, Codex レビュー
- allowed-tools:
- Bash(codex:*) Bash(git diff:*) Bash(git log:*) Bash(git stash:*) Bash(git status:*) Bash(cargo test:*) Bash(npm test:*) Bash(npm run test:*) Bash(bun test:*) Bash(pnpm test:*) Bash(yarn test:*) Bash(make test:*) Bash(which:*) Read Edit Skill Task AskUserQuestion
- model:
- opus
- argument-hint:
- [target scope]
/polish - Codex Review + Cleanup
Codex reviews the diff, then critic-audit puts the findings through an adversarial challenge. Fix the survivors, then clean up with /simplify and enhancer-code. Do not stop at a findings report; apply the fixes directly. For an internal-reviewer audit, use /audit.
Input
$ARGUMENTS may contain the target scope. If empty, analyze git diff HEAD (staged and unstaged changes).
Phase 1: Codex review
Run an external Codex review to produce findings. If which codex fails, skip to Phase 3.
Run codex review "Review for logic, architecture, data flow, and code simplicity (flag over-complexity and unnecessary indirection)". In codex 0.141.0 the scope flags (--uncommitted / --base / --commit) are mutually exclusive with the PROMPT argument, so pass no scope flag when sending the simplicity-lens PROMPT. Codex reads git status itself and reviews the staged, unstaged, and untracked changes. Omitting the PROMPT falls back to Codex's default review and drops the simplicity lens, so always pass it.
Phase 2: Adversarial challenge + fix
critic-audit filters Codex false positives, then the surviving findings are fixed directly.
- Map each finding to the
critic-auditinput schema (critic-audit§ Input) and spawnTask(subagent_type: "critic-audit")once with the full list. Wait for the per-finding verdict - Triage each finding by its verdict, per the table below
- From the survivors, keep Codex P1/P2. Drop Phase 3 territory (cleanup scope) and any fix touching files outside
git diffscope - Fix the survivors high severity first, then detect the test command and validate.
git stashany fix that breaks tests
| Verdict | Action | | ------------- | ------------------------------------------- | | confirmed | Keep as a fix candidate | | disputed | Drop | | downgraded | Apply the lowered severity | | needs_context | Surface to the user with a one-line summary |
Phase 3: Code cleanup
Two cleanup mutators run in sequence on the post-Phase-2 diff, then validate. Both apply fixes directly without a critic-audit challenge, since neither hunts bugs.
- Run
Skill("simplify")on the current diff for a cleanup-only pass (reuse, simplification, efficiency, altitude). If it rejects a no-arg invocation, pass the diff scope Phase 1 detected - Spawn
enhancer-codeviaTask(subagent_type: "enhancer-code")to remove AI slop and apply simplification rules, then audit tests. Run it after/simplifyso its preservation rule (when in doubt, keep) takes priority over/simplify's edits - Detect the test command and validate.
git stashthe cleanup edits on failure
Output
Report what each phase did. Codex findings (Phase 1), critic-audit verdict counts and fixes applied with skip reasons (Phase 2), and /simplify plus enhancer-code edits with file:line and the test validation result (Phase 3).
Error Handling
| Error | Action |
| ------------------------------------- | ---------------------------------------------- |
| No changes in diff | Report that there is nothing to polish |
| codex review fails | Skip to Phase 3 with no findings |
| critic-audit fails or returns empty | Keep all findings as confirmed, proceed to fix |
| /simplify unavailable or fails | Skip to enhancer-code |
Related Skills
thkt/use-cli-herdr
tools
VerifiedTrustedCommunity
Delegate implementation to codex (coder) via the herdr-agentchat plugin and drive a two-pane conversation to completion.
11SKILL.mdUpdated Jul 27, 2026
thkt/scribe
development
VerifiedTrustedCommunity
Extract recurring patterns from past closed PRs/issues and the research findings in workspace/research/, verify them against the latest code, and propose them to docs/wiki/ via PR.
11SKILL.mdUpdated Jul 20, 2026
thkt/dr
development
VerifiedTrustedCommunity
Create Decision Records (DR) in MADR v4 format with auto-numbering.
11SKILL.mdUpdated Jul 15, 2026
thkt/shake
development
VerifiedTrustedCommunity
Detect flaky tests by shaking them — repeated runs under varied order, parallelism, and seed — plus a static smell scan that flags latent flakiness in tests that currently pass. Classify each target as confirmed-flaky, latent-flaky, or stable and fix the root cause without weakening the test. Do NOT use to fix a confirmed single bug (use /fix) or for static-only code review (use /audit).
11SKILL.mdUpdated Jul 13, 2026