claude-power-user

Periodic, repo-agnostic doctor for AI-context files (CLAUDE.md, AGENTS.md, .cursorrules, .github/copilot-instructions.md). Verifies every concrete claim against the current repo, lints against an opinionated quality spec, surfaces newly-added subsystems the doc misses, and proposes the minimum edits to restore health. Conservative pruning. No-op when files are already healthy — safe to run on a cron.

Refine GitHub issues by creating WHAT/WHY/HOW sections and preparing them for implementation. Adds issues to GitHub Projects V2, sets status to "Ready", and assigns the current iteration.

Transform a GitHub issue into a production-ready pull request with automated PR review. Executes full workflow including refinement check, code generation, RL loop, testing, PR creation, automated review, feedback application, and re-verification before marking ready for human review.

Shared skill for GitHub Projects V2 API operations. Includes GraphQL mutations for adding issues, updating status, and managing iterations.

Transform ambiguous GitHub issues into refined specifications with WHAT, WHY, and HOW sections. Use before implementing features or fixes when issue lacks clear requirements.

Generate production-quality code following acceptance criteria from refined issue, with comprehensive tests and documentation. Use after issue refinement (Phase 1) is complete.

Iteratively improve code through CI/CD feedback until all quality gates pass. Use after code generation (Phase 2) to perfect implementation through automated testing and review.

Create comprehensive PR with automated review and feedback application. Creates a draft PR, runs automated review, fixes critical/high findings, re-verifies CI, and marks ready for human review.

Critical security-focused PR review for GitHub Actions CI. Only posts feedback when issues are found. Prefers inline comments over summaries.

Shared quality gates all phases must pass before proceeding. Single source of truth for pre-commit, tests, code quality, security, performance, documentation, CI, and acceptance criteria.

Architecture auditor for code quality, patterns, and technical debt. Use for full codebase audits, reviewing recent changes, or analyzing specific features. Provides actionable insights with impact assessment.

Identifies and updates outdated documentation and AI context files (CLAUDE.md, README.md, docs/). Use after feature work, PRs, or periodically to keep contexts accurate for AI agents. Surgical updates only — no unnecessary changes.

Development agent for implementing features, fixing bugs, refactoring, and all engineering work. Context-aware — reads project structure, docs, architecture, and conventions before writing code. Use for any development task.

Critical PR reviewer that checks code quality, security, performance, architecture, UI/UX, documentation, observability, and existing feedback. Use for reviewing pull requests before merge. Avoids nit-picking, focuses on issues that matter.

Security auditor for identifying vulnerabilities, misconfigurations, and unsafe patterns. Use for full security audits, reviewing recent changes for security regressions, or auditing specific attack surfaces. Produces triaged findings with severity, exploitability, and remediation guidance.

Use codenav CLI tool for efficient code navigation to explore function dependencies, call graphs, and code relationships. CRITICAL for token efficiency - always use codenav BEFORE grep/read when navigating codebases. Use when tracing function calls, finding dependencies, understanding architecture, or locating code patterns.