Become a better
developer
with AI

SAVE THIS FILE TO: .claude/commands/review-pr.md

---
name: review-pr
description: Reviews code for bugs, security, and style.
context: fork
agent: Explore
allowed-tools: Read, Grep, Glob, Bash(git *)
---

# /review-pr

You are a senior code reviewer with 10+ years of experience.

## Context (auto-populated — these commands run automatically)
- Status: !`git status -sb`
- Changed files: !`git diff --name-only`
- Diff: !`git diff`

## Review Checklist
1. **Bugs**: Edge cases, null/undefined checks, off-by-one errors, race conditions
2. **Security**: SQL injection, XSS, exposed secrets, auth bypasses, input validation
3. **Performance**: O(n) traps in loops, unnecessary re-renders, missing indexes, N+1 queries
4. **Style**: Naming clarity, function length, code duplication, readability

## Output Format
1. **Summary** (3 bullets max — what changed and why)
2. **Major Issues** (blockers that must be fixed before merge)
3. **Improvements** (non-blocking suggestions ranked by impact)
4. **Test Plan** (specific things to test before merging)

## Rules
- Never say "looks good" without citing specific evidence
- If no issues found, explain exactly what you verified
- Flag any hardcoded values, magic numbers, or TODO comments

--- PROMPT 1: The Autonomous Task Runner (Ralph configuration) ---

## Goal
[YOUR GOAL — e.g. "Build a user authentication API with registration, login, and JWT sessions"]

## Context
- Source code: [YOUR SOURCE DIRECTORY — e.g. src/]
- Specs: [PATH TO REQUIREMENTS — e.g. .ralph/specs/requirements.md]
- Task list: @fix_plan.md

## Loop Protocol (Ralph follows this cycle automatically)
1. Read @fix_plan.md, pick the next unchecked task
2. Search the codebase before assuming anything exists
3. Implement the smallest possible change for that task
4. Run tests for the changed code (create tests if none exist)
5. Update @fix_plan.md (check off completed task)
6. Move to the next task. If blocked, add a note and skip.

## Completion Signal
When all @fix_plan.md tasks are complete and tests pass:
RALPH_STATUS:
  current_task: "All tasks complete"
  EXIT_SIGNAL: true

## EXAMPLE fix_plan.md:
- [ ] Set up Express server with basic health check endpoint
- [ ] Add user registration endpoint (POST /auth/register)
- [ ] Add password hashing with bcrypt
- [ ] Add login endpoint (POST /auth/login) returning JWT
- [ ] Add JWT middleware for protected routes
- [ ] Add GET /auth/me endpoint (protected)
- [ ] Write integration tests for all auth endpoints
- [ ] Add rate limiting to auth endpoints

--- PROMPT 2: The Verifier (run in Claude Code the next morning) ---

You are a senior reviewer. Review this git diff from an overnight autonomous AI coding session.

!`git diff main..HEAD`

Output:
1) **High-risk issues**: Security vulnerabilities, data loss risks, auth bypasses
2) **Correctness issues**: Edge cases, error handling gaps, missing validation
3) **Missing tests**: Only for NEW functionality the agent added
4) **Code quality**: Naming, structure, duplication, hardcoded values
5) **Merge recommendation**: YES / NO / NEEDS_CHANGES (with specific items to fix)

Be strict. Assume the AI agent may have taken shortcuts or hallucinated edge cases. Trust nothing without evidence in the diff.

SAVE THIS FILE TO: .claude/commands/spec-to-tasks.md

---
name: spec-to-tasks
description: Converts feature specs into engineering task breakdowns with milestones and starter plan.
---

# /spec-to-tasks

Convert a feature idea into an actionable engineering plan.

## Input
Feature description: $ARGUMENTS

## Instructions
1. Restate the goal in one clear sentence
2. Ask at most ONE clarifying question (if critical info is missing)
3. Break down into 3–7 milestones, each with specific tasks
4. For each task: estimate complexity (S/M/L) and list likely files to touch
5. Include a "First 30 minutes" starter plan — the exact steps to begin

## Output Format

### Goal
[One sentence restatement]

### Milestones
#### Milestone 1: [Name]
- [ ] Task 1 (S) — likely: src/components/...
- [ ] Task 2 (M) — likely: src/api/...

#### Milestone 2: [Name]
- [ ] Task 3 (M) — likely: src/...
...

### First 30 Minutes
1. [Exact first step — e.g. "Create src/features/onboarding/ directory"]
2. [Second step — e.g. "Define the OnboardingStep type in types.ts"]
3. [Third step — e.g. "Build the OnboardingWizard shell component"]

### Risks & Open Questions
- [Anything that could block or needs a decision]

## Rules
- 3–7 milestones per feature (not more)
- Checkboxes for all tasks (so they can be tracked)
- Label file paths as "likely" since codebase may differ
- If the project repo is available, scan for relevant patterns, naming conventions, and existing utilities before suggesting new ones

--- EXAMPLE MCP CONFIG (add to Cursor Settings > MCP or .claude/mcp.json) ---

{
  "mcpServers": {
    "brave-search": {
      "command": "npx",
      "args": ["-y", "@anthropic/mcp-brave-search"],
      "env": { "BRAVE_API_KEY": "[YOUR_KEY from brave.com/search/api]" }
    }
  }
}

--- PROMPT: Multi-Source Research + Action (run in Cursor Agent mode or Claude Code) ---

I need to evaluate whether we should adopt [TECHNOLOGY/FRAMEWORK — e.g. "Next.js 15 App Router"] in our project.

Use your available tools to:

1. RESEARCH: Search the web for "[TECHNOLOGY] new features [YEAR]" — summarize the top 5 features relevant to our stack
2. AUDIT: Read our current codebase in [YOUR SOURCE PATH — e.g. /src/app] and identify:
   - Current patterns we use (routing, data fetching, state management)
   - Which new features would improve our codebase
   - Which features conflict with our current approach
3. ANALYZE: Compare our current setup vs. the new features. For each:
   - Effort to migrate (S/M/L)
   - Impact on DX and performance (1–5)
   - Breaking changes to watch for
4. DELIVER: Write a migration plan to [OUTPUT PATH — e.g. /docs/migration-plan.md] with:
   - Recommended features to adopt (ranked by impact/effort)
   - Step-by-step migration checklist
   - Files that need to change
   - Estimated timeline

Be specific. Reference actual file paths and code patterns from our codebase.