Skills, agents, and specs for autonomous coding workflows
This repository also configures its own agents. See what agent-kit tells them →
Skill Claude CodeCodex
Plan an OpenSpec change and sync it to the MCP Task Hub as tiered, chunked tasks. Use when the user wants to plan/queue work for the agent drain loop ("hub-plan", "sync tasks to the hub", "queue up this change"). Spec artifacts always come from the official OpenSpec (opsx) tooling — this skill layers reconciliation…
Skill Claude CodeCodex
Show the MCP Task Hub queue — counts, blocked tasks with reasons, stale claims — grouped by project and change. Use when the user asks "hub status", "what's in the queue", "any blocked tasks", or after a drain run. The browser view is http://taskhub.local/ui (LAN).
Skill Claude CodeCodex
Attach AI observability notes to git commits using refs/notes/agent-log. Load this skill when you need to record why a change was made after committing.
Skill Claude CodeCodex
The step-by-step protocol every hub-worker tier follows to execute exactly one MCP Task Hub task in a lane worktree — fetch, claim, read narrowly, implement, gates, zero-bloat, commit locally, report the commit range. Loaded by hub-worker, hub-worker-haiku and hub-worker-opus.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: