claude-toolkit

This skill should be used when the user asks to "orchestrate tasks", "run parallel agents", "wave orchestration", "iterative agent loops", "swarm agents", "coordinate multiple agents", or needs guidance on decomposing complex tasks into parallel workstreams with iterative refinement.

Nightly maintenance for the compound loop — review each project STATE.md, promote only verified lessons to durable rules, keep MEMORY.md under budget, and log a short digest. Conservative and local only — never edits source, never pushes, never messages anyone. Schedule it to run nightly (see Scheduling).

Use when finishing a non-trivial task, fixing a bug, completing a feature, or wrapping up a session and you want the work to make the system smarter next time. Runs the compound loop — consult memory, do the work, verify with an independent subagent (max 3 cycles), then distill only CONFIRMED lessons into STATE.md and memory. Triggers on "compound", "make this stick", "close the loop", or end-of-task wrap-up.