domains / general / review / review

review

Skill card

combined

The four copies share the name 'review' but define four unrelated procedures. dzhng/skills is a closeout pass that sequences refactor-clean, code-review, and write-docs for a finished code change. davepoon/buildwithclaude is a cross-AI peer review of phase plans that invokes external CLIs and writes REVIEWS.md. LeoYeAI/openclaw-master-skills is an audit of Claude Code auto-memory for promotion, staleness, and consolidation. brycewang-stanford/Auto-Empirical-Research-Skills is a research-paper review router that dispatches critic agents for peer review, stress tests, and code replication. No copy is a reproduction of another; each contains material steps and constraints the others lack, so crowning one would hide real alternatives.

Repositories

Other copies

Combined skill

---
name: review
description: Unified review skill that routes to the appropriate review procedure based on the target and flags.
argument-hint: "[target] [--code] [--plans] [--memory] [--paper] [--peer [journal]] [--stress [journal]] [--methods] [--proofread] [--replicate [lang]] [--all]"
---

# Review

Routes to the appropriate review procedure based on the target and flags.

## Routing Logic

- `--code` or a code file path → **Code Closeout Review** (dzhng/skills procedure)
- `--plans` or a phase plan path → **Cross-AI Plan Review** (davepoon/buildwithclaude procedure)
- `--memory` → **Auto-Memory Audit** (LeoYeAI/openclaw-master-skills procedure)
- `--paper` or a `.tex` file path → **Research Paper Review** (brycewang-stanford procedure)
- `--all` or no target → Run all applicable reviews

## Code Closeout Review

1. **Scope.** Fix the diff under review and confirm it's substantive; skip for trivial edits.
2. **Shape.** Run refactor-clean first, while restructuring is cheap.
3. **Diff.** Run code-review on the settled shape. A finding that forces a structural change sends you back to step 2.
4. **Docs.** Run write-docs. Trace the link chain from the root README down to each doc; every link must resolve.
5. **Report.** One verdict across all three passes — what each found, what you changed, what you left and why.

## Cross-AI Plan Review

1. **Detect CLIs.** Identify available external AI CLIs (Gemini, Claude, Codex, OpenCode, Qwen, Cursor).
2. **Build review prompt.** Construct the review prompt for the phase plan.
3. **Invoke each CLI.** Run each selected CLI independently.
4. **Collect responses.** Gather all reviewer feedback.
5. **Write REVIEWS.md.** Produce a structured REVIEWS.md with per-reviewer feedback.

## Auto-Memory Audit

1. **Locate memory directory.** Find the project's auto-memory directory.
2. **Read and analyze MEMORY.md.** Count lines, check against the 200-line limit, analyze for recurrence, staleness, consolidation, and promotion candidates.
3. **Read topic files.** Cross-reference with MEMORY.md for duplicates.
4. **Cross-reference with CLAUDE.md.** Check for duplicates, conflicts, and promotion candidates.
5. **Generate report.** Output memory health, promotion candidates, stale entries, consolidation groups, conflicts, and recommendations.

## Research Paper Review

1. **Auto-detect by file type.** Route to comprehensive review, code review, or talk review.
2. **Apply explicit flags.** Override auto-detect with `--peer`, `--stress`, `--methods`, `--proofread`, `--code`, `--replicate`, or `--all`.
3. **Execute the selected mode.** Dispatch the appropriate critic agents and produce the structured report.

## Conflicts

- dzhng/skills sequences refactor-clean, code-review, and write-docs for a finished code change; loops back when a later pass reopens an earlier one.
- davepoon/buildwithclaude invokes external AI CLIs to review phase plans and writes REVIEWS.md.
- LeoYeAI/openclaw-master-skills audits Claude Code auto-memory for promotion, staleness, and consolidation.
- brycewang-stanford/Auto-Empirical-Research-Skills routes to critic agents for peer review, stress tests, and code replication of research papers.