Claude Skill MartBrowse skillsWhat is a Skill?
 
← Back to list

review-work (5-Agent Parallel Review Orchestrator)

Runs five specialized sub-agents in parallel — goal verification, hands-on QA, code quality, security, and context mining — and only passes the review when all five pass.

Dev & CodingAdvanced★ 67,531⑂ 5,507AI score 8/10Last updated: Aug 9, 2026

What it does

Right before a PR handoff, this skill fans the same change set out to five specialized reviewers running concurrently:

  1. Goal & constraint verification — decomposes the original request into explicit and implied sub-requirements, marks each ACHIEVED/MISSED/PARTIAL with code evidence, and flags over-engineering.
  2. Hands-on QA execution — brainstorms 15–30 scenarios, augments them, converts them into a prioritized task list, then actually runs the app (web, CLI, API, library) and records PASS/FAIL with evidence.
  3. Code quality review — ten dimensions from correctness and pattern consistency to type safety, performance, test quality and tech debt, graded CRITICAL/MAJOR/MINOR/NITPICK.
  4. Security audit — injection, auth/authz, hardcoded secrets, data exposure, path traversal, dependency CVEs, error leakage.
  5. Context mining — digs through git log/blame, GitHub issues and past PR review comments, plus Slack/Notion docs to surface background the implementer missed.

It also enforces reviewing branches from a dedicated git worktree, and records silent lanes as INCONCLUSIVE so they are never counted as approval.

Who it's for

  • Developers who want a rigorous self-review gate before opening a PR
  • Small teams or solo devs without a second reviewer
  • Tech leads who need a safety net over AI-generated code
  • Users on Codex/OpenCode-style harnesses with parallel subagent support

Examples

  • Pre-PR gate: say "review my work"; it auto-collects git diff main...HEAD, launches all five lanes in the background, and returns a table-formatted PASSED/FAILED report with a prioritized fix list.
  • Payment feature audit: the security lane catches a hardcoded API key and internal details leaking in error responses, while the QA lane actually exercises declined-card and timeout paths.
  • Missed-requirement catch: the context-mining lane finds GitHub issue #42 and a previously reverted commit, reporting a BLOCKING finding because the new implementation repeats the same mistake.

· · · Install guide · · ·

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the packages/shared-skills/skills/review-work folder from the GitHub repo code-yeongyu/oh-my-openagent into my ~/.claude/skills/review-work/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/code-yeongyu/oh-my-openagent.git /tmp/omo && mkdir -p ~/.claude/skills/review-work && cp -r /tmp/omo/packages/shared-skills/skills/review-work/* ~/.claude/skills/review-work/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal and clone the repo into a temp folder: git clone https://github.com/code-yeongyu/oh-my-openagent.git /tmp/omo
  2. Create the skill directory: mkdir -p ~/.claude/skills/review-work
  3. Copy the skill files: cp -r /tmp/omo/packages/shared-skills/skills/review-work/* ~/.claude/skills/review-work/
  4. Restart Claude Code and confirm review-work appears in your skill list.
  5. From a repo with the branch you want reviewed checked out, ask: "review my work" / "post-implementation review".
  6. Prerequisites: git is required; run gh auth login if you want GitHub issue/PR mining, and install Playwright or a browser tool for web-app QA.
  7. Note: if your environment lacks parallel subagent tools, ask the assistant to run the five lane prompts sequentially — same coverage, just slower.
View source on GitHub ↗License: NOASSERTION