A builder/critic loop, a fleet of scheduled cloud agents with a watchdog, a CLAUDE.md keeper, a prompt-corpus miner, and a todo list that spans sessions.
A Claude Code plugin that runs a builder/critic improvement loop: fresh critic agents review each round against a frozen goal until the work passes.
A Claude Code skill that deploys a fleet of scheduled cloud agents from a git-committed manifest, with a daily watchdog that audits the fleet and reports in.
A Claude Code skill that keeps project CLAUDE.md files true and lean: distills non-obvious session learnings into the file and prunes what went stale.
Captures every Claude Code prompt into a durable local corpus, then mines it weekly for the skills and CLAUDE.md rules you should have written already.
A global todo list for Claude Code where every item links back to the unfinished conversation that created it.