kimi-lane
Skill1 starsChanged 4 days agoMITView on GitHub
Run a job on Kimi models through the Kimi Code CLI (kimi -p): cheap execution of easy, testable work in its own git worktree, or a quick extra opinion. Use when the user asks for Kimi, when other lanes hit their limits, or when the orchestra routing table picks Kimi. Needs the orchestra skill's scripts.
In this repository4 other files it gives its agents.
- chatgpt-laneSkill
- councilSkill
- deepseek-laneSkill
- orchestraSkill
The file
Files
Kimi lane (Kimi Code CLI)
Models
| Alias | Use |
| kimi-code/k3.1-pro | Default. Execution from explicit steps, opinions |
| kimi-code/k3.1-pro-highspeed | Same model, faster; needs a higher Kimi plan (else status=limit) |
kimi reads its model aliases from its config file. List them with mmo-preflight.
Run
git worktree add ../wt-<task> -b lane/<task>
../orchestra/scripts/mmo-run kimi --mode rw -C ../wt-<task> --brief brief.md --out out/<task>.md
../orchestra/scripts/mmo-run kimi --mode ro --brief q.md --out out/kimi.mdkimi -p applies edits without asking, so the runner refuses --mode rw
outside a linked worktree. The runner reads the last assistant message from Kimi's JSON stream.
Brief
Kimi does best with very explicit briefs: numbered steps, exact file names, the exact test command, and a short list of things it must not do. Give it one job per run.
After it finishes
Same gates as every lane: status ok, re-run the check, read the diff, merge. On limit move to the
chatgpt-lane or deepseek-lane. On auth ask the user to log in to Kimi Code again.
Read-only lanes
Kimi Code has no read-only sandbox: kimi -p may edit files even when told not to. In --mode ro the
runner gives it a throwaway detached worktree (or an empty folder outside git), discards whatever it wrote,
and marks the header ro_violation=yes if it tried.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.