home / skills / phrazzld / claude-config / review-and-fix

review-and-fix skill

safe

/skills/review-and-fix

This skill orchestrates a full review-to-pr workflow, analyzing code, applying fixes, and enforcing quality gates to ship PR-ready Python code.

npx playbooks add skill phrazzld/claude-config --skill review-and-fix

Review the files below or copy the command above to add this skill to your agents.

Files (1)

SKILL.md

1.2 KB

---
name: review-and-fix
description: |
  Full code review, fix, quality, PR workflow.
  Chains review-branch, address-review, check-quality, and pr.
  Use when: code complete and ready for PR, want comprehensive review before shipping.
effort: high
---

# /review-and-fix

Code complete to PR ready in one command.

## Role

Thin orchestrator chaining review primitives.

## Objective

Take current branch from "code complete" to "PR ready" with comprehensive review, fixes, and quality gates.

## Latitude

- Run full review pipeline or skip steps with flags
- Loop back for re-review if Critical items found
- Create GitHub issues for out-of-scope findings

## Usage

```
/review-and-fix              # Full flow
/review-and-fix no-pr        # Stop after fixes
/review-and-fix verify       # Re-review after fixes
```

## Workflow

1. **Review** — `/review-branch` (~12 reviewers, parallel)
2. **Fix** — `/address-review` (TDD: failing test, fix, passing test, commit)
3. **Quality** — `pnpm typecheck && pnpm lint && pnpm test`
4. **Re-review** — If `verify` flag or Critical items found, loop to step 1
5. **Ship** — `/pr` (unless `no-pr`)

## Output

Summary: reviewers consulted, findings by severity, fixes applied, issues created, quality gates, PR URL.

Overview

This skill automates a full code review-to-PR workflow to get a branch from code-complete to PR-ready in one command. It orchestrates parallel reviews, guided fixes with test-driven steps, quality gates, and final PR creation while surfacing findings and actions taken. Use it when you want a comprehensive, repeatable pre-merge process.

How this skill works

The skill chains four primitives: review-branch, address-review, check-quality, and pr. It runs parallel reviewers to collect findings, applies fixes using a test-first loop, runs typechecks/lint/tests, and optionally re-runs reviews for critical items before creating the PR. Outputs a concise summary with reviewers consulted, severity breakdown, fixes, issues opened, quality results, and PR URL.

When to use it

Branch is code-complete and you want one-command readiness for PR
You need a thorough multi-reviewer assessment before shipping
You want fixes applied following a TDD-style loop (failing test → fix → passing test)
You need automated quality gates (typecheck, lint, tests) enforced
You want to open issues for out-of-scope findings and re-run verification after fixes

Best practices

Run the full flow without flags on feature-complete branches to catch regressions early
Keep commits small and focused during the address-review step to make reviewable changes
Enable the verify flag when fixes touch critical logic so the review loop re-runs automatically
Treat created issues as backlog items for long-running or architectural concerns
Review the output summary and listed fixes before final PR creation to ensure intent matches changes

Example use cases

Finalize a feature branch and produce a high-confidence PR in a single command
Post-implementation pass to convert reviewer feedback into focused commits with passing tests
Enforce CI quality gates locally before opening a PR to reduce CI cycle time
Loop verification when a security or critical bug is discovered during review
Automatically create GitHub issues for design concerns identified during review

FAQ

Can I skip PR creation?

Yes. Use the no-pr flag to stop after fixes and quality checks so you can inspect changes before opening a PR.

What triggers a re-review loop?

Use the verify flag to force re-review after fixes. The skill also auto-loops if reviewers report Critical severity findings that require another evaluation.