codexstar69 Code Perfection — Autonomous Refactoring Agent
Use this skill when the user asks to refactor, optimize, clean up, or audit code for quality issues. The skill provides enforced resolution loops that cannot exit until all issues are resolved or explicitly deferred.
Tools Provided
| Tool | Purpose |
|---|---|
codeperfect_verify |
8-point verification checklist (compiles, tests, no any, no secrets, no dead code, no scope creep) |
codeperfect_loop |
Resolution loop — issue ledger, lock, revert-on-failure, auto-commit, auto-defer |
codeperfect_audit |
Domain-scoped audit state — triage, one-domain-at-a-time, boundary analysis, merge-findings |
codeperfect_triage |
Structural recon — risk-tier classification without reading code |
Commands
| Command | Description |
|---|---|
/code-perfection <path> |
Run the resolution loop on a target |
/code-perfection:audit <path> |
Full tiered codebase audit |
/code-perfection:verify |
Run verification checklist |
Code Standards
Read references/agents.md for the full code standards the agent must follow during refactoring.
Resolution Loop Protocol
Read references/resolution-loop.md for the loop protocol: SCAN → PLAN → FIX → VERIFY → DECIDE → PROGRESS CHECK.
Audit Strategy
Read references/audit.md for the tiered audit approach: Tier 0 (triage) → Tier 1 (domain audits) → Tier 2 (boundary audits) → Tier 3 (merge & report).
Parallel vs Sequential
Read references/parallel-vs-sequential.md for when to use parallel vs sequential scanning.