skills

Run a large, correctness-critical build (a port, rewrite, protocol/serializer, or migration) as a team of AI agents — a persistent coordinator that never writes bulk code dispatches one implementation agent per milestone, and every PR passes a merge gate whose adversarial half finds the bugs green tests miss. Use when the user wants to port or rewrite a big codebase, run a multi-milestone project with subagents, set up review gates for agent-authored PRs, or coordinate a fleet of coding agents against a reference implementation.

Analyze your own AI coding-assistant session history (Claude Code AND OpenAI Codex) to discover recurring workflows and propose new skill candidates (global or project-scoped), plus surface usage insights. Use when the user wants to mine/analyze their logs/transcripts/history for skill ideas, find repeated tasks worth automating, or audit how they use their AI tools.