| name | kimi-cli-skill |
| description | Reference guide for integrating and operating Kimi Code CLI in Ode, with focus on prompt mode, sessions, and automation-safe defaults. |
What I do
- Provide Kimi CLI integration guidance for Ode agent providers.
- Document reliable non-interactive invocation patterns using
-p/--prompt.
- Explain session reuse (
--session) and JSONL output parsing for current Kimi Code CLI.
- Call out automation implications for prompt mode.
When to use me
Use this when adding or debugging the kimi provider in Ode, especially command construction, streaming output parsing, or script/CI behavior.
Ask clarifying questions if you need model/auth-specific setup beyond CLI invocation.
Recommended invocation pattern
- New session:
kimi --output-format stream-json -p <prompt> from the target working directory.
- Resume session:
kimi --output-format stream-json --session <session_...> -p <prompt>.
- Kimi Code 0.13.x does not support the older
--print or --work-dir flags.
- The CLI stores sessions in
~/.kimi-code/session_index.jsonl; new session IDs use the session_... shape.
Integration notes for Ode
- Treat prompt mode as the automation surface; add
--auto/--yolo only when the target CLI version requires it for tool approval.
- Parse stdout as JSONL message stream (
assistant and optional tool messages).
- Keep provider model input hidden in UI unless explicitly needed.
- Keep session IDs stable per Slack thread and rotate when environment changes.
Sources
