shield_lock
JinDaGe
← Back to skills
extension
Category: Development & EngineeringNo API key required

pr-review-standards

Use when creating PRs to enforce code quality standards. Automated detection of anti-patterns, security issues, and code smells. Python 3.8+

personAuthor: jakexiaohubgithub
open_in_newRepositorydownloadDownload package

Skill: PR Review Standards

Purpose

Enforce code quality before human review. Transform the agent from a "coder" into a "reviewer" by catching common anti-patterns and violations.

1. Negative Knowledge (Anti-Patterns)

CRITICAL: Fail the review if any of these are present.

| Pattern | Why It Fails | Correction | | :--- | :--- | :--- | | console.log in commits | Pollutes production logs | Use logger.debug() or remove | | Hardcoded secrets | Security violation | Use process.env or secrets manager | | Big PRs (>500 lines) | Unreviewable, high risk | Split into feature branches | | Direct DB access in routes | Violates separation of concerns | Use service layer | | Missing tests for new code | No verification of behavior | Add unit/integration tests | | Task marker comments (TO-DO/FIX-ME) in production code | Technical debt indicator | Resolve or create issue | | Commented-out code blocks | Code smell, use version control | Delete and rely on git history | | Inconsistent formatting | Readability issues | Run formatter before commit |

2. Verified Procedure

Pre-Review Checks

  1. Scan Diff: Run git diff main...HEAD to see all changes
  2. Analyze: Check against the "Negative Knowledge" table above
  3. Validate: Run the static analysis script: 
    python .claude/skills/pr-review-standards/scripts/analyze_diff.py
  4. Report: Generate a markdown summary of blocking issues

Review Criteria

  • Security: No hardcoded credentials, API keys, or tokens
  • Quality: No debug statements, commented code, or task markers
  • Size: PR should be <500 lines (excluding generated files)
  • Tests: New features must include tests
  • Documentation: Public APIs must have docstrings

Execution Flow

1. Agent completes feature work
2. Agent runs self-review using this skill
3. If blocking issues found:
   - Fix immediately
   - Re-run validation
4. If clean:
   - Commit changes
   - Push to branch
   - Create PR

3. Zero-Context Scripts

analyze_diff.py

Located at: .claude/skills/pr-review-standards/scripts/analyze_diff.py

Purpose: Automated static analysis of git diff to detect anti-patterns.

Usage:

python .claude/skills/pr-review-standards/scripts/analyze_diff.py [--base=main]

Returns:

  • Exit code 0: No blocking issues
  • Exit code 1: Blocking issues found
  • JSON report of all findings

4. Failed Attempts (Negative Knowledge Evolution)

❌ Attempt: Allow console.log in development

Context: Tried to permit console.log if guarded by environment checks Failure: Still leaked to production due to misconfigured env vars Learning: Zero tolerance for console.log in commits

❌ Attempt: Auto-fix all issues

Context: Script attempted to automatically fix detected issues Failure: Introduced bugs by removing contextually-important code Learning: Report issues, require manual review and fixing

5. Governance

  • Token Budget: ~300 lines (within 500 limit)
  • Dependencies: Git, Python 3.8+
  • Maintenance: Update "Negative Knowledge" table as new patterns emerge
  • Verification Date: 2026-01-01