# Kimi Source Notes

Sources checked on 2026-05-16:

- Kimi Code CLI print mode: https://moonshotai.github.io/kimi-cli/en/customization/print-mode.html
- Kimi Code CLI command reference: https://www.kimi.com/code/docs/en/kimi-code-cli/reference/kimi-command.html
- Kimi Code CLI skills: https://moonshotai.github.io/kimi-cli/en/customization/skills.html
- Kimi Code CLI agents and subagents: https://moonshotai.github.io/kimi-cli/en/customization/agents.html
- Kimi prompt best practices: https://platform.kimi.ai/docs/guide/prompt-best-practice
- Kimi Websites overview: https://www.kimi.com/help/websites/websites-overview

Key verified points:

- `kimi --print` runs Kimi Code CLI non-interactively and exits after executing instructions.
- Non-interactive input can be supplied with `-p` / `--prompt`, `-c` / `--command`, or stdin.
- `--quiet` is shorthand for `--print --output-format text --final-message-only`.
- `--input-format=stream-json` and `--output-format=stream-json` use JSONL message streams.
- Print mode returns exit code `0` for success, `1` for non-retryable failures, and `75` for retryable transient failures.
- `--skills-dir PATH` can expose additional skills; Kimi also discovers user and project skills.
- In local `kimi --help` for version 1.44.0, `--skills-dir` is described as repeatable custom skills directories that override default discovery, so include every needed skill directory when using it.
- Kimi prompting guidance emphasizes clear instructions, role assignment, delimiters, explicit steps, examples, target length, reference text, and decomposing complex tasks.
- Kimi Websites documentation emphasizes goal, audience, design style, visual inputs such as screenshots or mockups, and multi-round revision for website generation.