itboy79

Multi-agent build orchestrator run like a kitchen brigade — one chef who writes the tickets, parallel cooks who fire them, a fresh-palate pass that tastes every diff. Use this skill whenever the user invokes /brigade (or the legacy /rehoboam), asks to orchestrate a build task across multiple agents, wants a feature split into parallel sub-tasks, mentions "chef and brigade", "tickets", "the pass", "refire", "orchestrator", "parallel executors", or asks to implement a non-trivial feature end-to-end (plan → build → review → integrate). Provider-agnostic - cooks can run on Anthropic, OpenAI-compatible APIs, or local Ollama models, configured per-repo.

Repo-wide knowledge engine — scans the entire repository (or project folder) and writes a Karpathy-style wiki-OS, numbered LLM-first markdown pages that become the project's single source of truth, then keeps it in sync forever. Use this skill whenever the user invokes /cerebro, asks to "scan the repo", "map the codebase", "read the whole project", create/update/sync a "wiki", "wiki-OS", "knowledge base" or "docs for LLMs", mentions "source of truth", "drift", onboarding an AI onto an unfamiliar codebase, or starts non-trivial development in a repo that has (or should have) a wiki/ directory. Model-agnostic — works with any LLM that can read and write files; no vendor tools, no APIs, no dependencies required.