kimi-cli

name	kimi-cli
description	Use when the user asks to run Kimi CLI in non-interactive mode or references Kimi for AI-assisted coding or agent tasks.

Kimi CLI Skill Guide

Before running Kimi, follow the shared logging and summary protocol in ../_shared/agent-cli.md.

Running a Task

1. Verify installation: command -v kimi
2. Select the mode required for the task; default to --print mode for non-interactive use. For implementation tasks, first use superpowers:using-git-worktrees to create a fresh git worktree, then run Kimi there with write-capable/auto-approve options when approved.
3. Always use the current host's user-question or approval mechanism before using -y, --yolo, --yes or --print modes (they auto-approve actions).
4. Assemble the command with the appropriate options:
   • -p, --prompt, --command <text> - User prompt to the agent (required for non-interactive)
   • -C, --continue - Continue previous session for working directory
   • -S, --session <id> - Session ID to resume
   • -m, --model <model> - LLM model to use
   • -y, --yolo, --yes - Automatically approve all actions (use with caution)
   • --print - Run in print mode (non-interactive, implies --yolo)
   • --quiet - Alias for --print with text output and final message only
   • -w, --work-dir <directory> - Working directory for the agent
   • --add-dir <directory> - Add additional directory to workspace scope
   • --agent <default|okabe> - Builtin agent specification
5. Do not suppress stderr. Capture stdout and stderr into full.md, and require the prompt to write summary.md.
6. Run the command, inspect the full log and summary, and summarize the outcome for the user.

Quick Reference

Use case	Command pattern
Non-interactive print mode	kimi --print --prompt "prompt"
Quiet mode (final msg only)	kimi --quiet --prompt "prompt"
Continue last session	kimi --print --continue --prompt "follow-up"
Resume specific session	kimi --print --session <id> --prompt "prompt"
Yolo mode in fresh worktree only (auto-approve)	kimi --print --yolo --prompt "prompt"
With specific model	kimi --print --model <model> --prompt "prompt"
Additional directory	kimi --print --add-dir /path --prompt "prompt"
Okabe agent	kimi --print --agent okabe --prompt "prompt"

Modes

Mode	Description	Use Case
--print	Non-interactive, auto-approves	Analysis only unless running inside a fresh worktree
--quiet	Non-interactive, only final output	Clean output for scripts
--yolo	Auto-approve all actions	Full access in a fresh worktree

Example Commands

# Non-interactive print mode; analysis in main tree, writes only in a fresh worktree
kimi --print --prompt "Refactor the utils module"

# Quiet mode (only final message)
kimi --quiet --prompt "Fix the bug in src/main.py"

# Continue last session
kimi --print --continue --prompt "What else can you improve?"

# With specific model and okabe agent
kimi --print --agent okabe --model moonshot-v1-auto --prompt "Analyze codebase"

# Capture stdout/stderr according to ../_shared/agent-cli.md

Following Up

• After every kimi command, immediately use the current host's user-question or approval mechanism to confirm next steps, collect clarifications, or decide whether to resume with --continue.
• When resuming, the session automatically uses the same model and agent from the original session.
• Restate the chosen model and agent when proposing follow-up actions.

Error Handling

• Stop and report failures whenever kimi --version or a kimi command exits non-zero; request direction before retrying.
• Before you use high-impact flags (--yolo, --yes, --print) ask the user for permission using the current host's user-question or approval mechanism unless it was already given.
• When output includes warnings or partial results, summarize them and ask how to adjust using the current host's user-question or approval mechanism.
• Always validate Kimi's output for security vulnerabilities (XSS, injection) before using.