codex-codebase-review

分类: 开发与编程 | 上传者: lploc94 | 下载: 0 | 版本: v1.0（最新）

Review entire codebases (50-500+ files) by chunking into modules, reviewing each chunk in a separate Codex session, then synthesizing cross-cutting findings. No runner changes needed.

更新日志: Source: GitHub https://github.com/lploc94/codex_skill

目录结构

当前层级: skill-packs/codex-review/skills/codex-codebase-review/

📁 references/
- 📄 output-format.md 1.3 KB
- 📄 prompts.md 3.3 KB
- 📄 workflow.md 10.9 KB
📄 SKILL.md 3.2 KB

SKILL.md

---
name: codex-codebase-review
description: Review entire codebases (50-500+ files) by chunking into modules, reviewing each chunk in a separate Codex session, then synthesizing cross-cutting findings. No runner changes needed.
---

# Codex Codebase Review

## Purpose
Review large codebases (50-500+ files) that exceed single-session context limits. Splits codebase into module-based chunks, reviews each in an independent Codex session, then Claude synthesizes cross-cutting findings across modules.

## Prerequisites
- Source files in working directory.
- `codex` CLI is installed and authenticated.
- `codex-review` skill pack is installed (`npx github:lploc94/codex_skill`).

## Runner

```bash
RUNNER="{{RUNNER_PATH}}"
```

## Workflow
1. **Collect inputs**: effort level, parallel factor, focus areas, output format (`markdown` default, `json`, `sarif`, or `both`).
2. **Discovery**: detect project type, list source files, identify module boundaries.
3. **Chunking**: group files into 500-2000 line chunks, present chunk plan.
4. **Review loop**: for each chunk — build prompt, `node "$RUNNER" start --format "$FORMAT"`, poll, parse ISSUE-{N}, propagate context.
5. **Cross-cutting analysis**: Claude synthesizes all chunk findings — inconsistencies, API contracts, DRY violations, integration, architecture.
6. **Validation** (effort >= high): feed CROSS-{N} findings to Codex for verification.
7. **Final report**: overview table, per-module findings, cross-cutting findings, action items.
8. **Cleanup**: stop ALL tracked STATE_DIRs — always runs regardless of outcome.

### Effort Level Guide
| Level    | Discovery        | Cross-cutting    | Validation   |
|----------|------------------|------------------|--------------|
| `low`    | Auto-detect only | Basic (2 cats)   | Skip         |
| `medium` | Auto + confirm   | Standard (3 cats)| Skip         |
| `high`   | Full + confirm   | Full (5 cats)    | 1 round      |
| `xhigh`  | Full + suggest   | Full + arch      | 2 rounds     |

### Output Format Guide
| Format     | Output Files                          | Best for                        |
|------------|---------------------------------------|---------------------------------|
| `markdown` | `review.txt` (human-readable)         | Default, interactive review     |
| `json`     | `review.txt` + `review.json`          | CI/CD integration, automation   |
| `sarif`    | `review.txt` + `review.sarif.json`    | IDE integration (VS Code, etc.) |
| `both`     | All above + `review.md` (rendered)    | Complete documentation          |

**Note**: `review.txt` is always written for backward compatibility. Each chunk produces separate output files in its STATE_DIR.

## Required References
- Detailed orchestration: `references/workflow.md`
- Prompt templates: `references/prompts.md`
- Output contract: `references/output-format.md`

## Rules
- Codex reviews only; it does not edit files.
- No cross-contamination between chunk sessions — each chunk is independent.
- Context propagation: only high/critical findings from prior chunks, capped at ~2000 tokens.
- Cleanup always runs — stop every tracked STATE_DIR regardless of outcome.
- Scope is full codebase only — for diff review use `/codex-impl-review`.

登录后下载/点赞/收藏 ❤ 16 | ★ 0

codex-codebase-review

目录结构

SKILL.md

举报内容

提示