Improve an Agent

Coding-agent workflow — a /slash-command your coding agent (Claude Code, Codex, others) runs while developing this repo. Invoke it by name (e.g. /improve-agent) or describe the task and it triggers automatically.

You are recursively improving a target agent autonomously. No user-supplied test cases — you derive your own probes from the agent's stated purpose (its INSTRUCTIONS), test the agent against them, judge the results, and iterate on agents/<slug>.py until the agent reliably does what its instructions say it does.

This is the autonomous half of the iteration loop. The user-driven half lives in extend-agent (add a tool, add a capability, refine the prompt, fix a specific bug). Use the extend-agent skill to change the agent; use this skill to harden it against its stated intent.

The platform is on http://localhost:8000 with hot-reload enabled (RUNTIME_ENV=dev), so edits to agents/<slug>.py are picked up by uvicorn within ~1s. No rebuild, no restart.

This is a single-pass loop. One pass usually takes 15-30 minutes depending on the agent's surface area. Re-run if behavior still drifts.

0. Preconditions

• Live container reachable: curl -sSf http://localhost:8000/health returns 200. If not, ask the user to docker compose up -d --build first. (docker compose ps is unreliable from worktrees or alternate clones — trust the health probe.)
• Live container is bound to this checkout — otherwise hot-reload won't see your edits: