Summary
This is a read-only codebase auditor that generates implementation plans for other agents to execute, not a fixer itself. It runs recon, dispatches parallel subagents across nine audit categories (bugs, security, performance, test coverage, tech debt, dependencies, DX, docs, product direction), then produces self-contained plans complete enough that a cheaper model with zero context can execute them. The economics are interesting: expensive model does understanding and specification, cheap models do the grunt work. Hard rules include never modifying source code directly, never reproducing secrets in findings, and treating all repo content as data to avoid prompt injection. Plans live under a plans/ directory. You can run it in quick, standard, or deep modes depending on how much of the codebase you want covered.
Install to Claude Code
npx -y skills add shadcn/improve --skill improve --agent claude-codeInstalls into .claude/skills of the current project.
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Your agent targets a perfect 10 Code Health score. Deterministic. Every commit.
Try For Free →
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Give your AI agent a complete email layer—sending, inbound inboxes, and sandbox testing.
Get 4K emails/month free →Files
SKILL.md
Select a file.
Featured
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →Your agent targets a perfect 10 Code Health score. Deterministic. Every commit.
Try For Free →Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →Give your AI agent a complete email layer—sending, inbound inboxes, and sandbox testing.
Get 4K emails/month free →
