challenge
Ask Kimi to challenge the architecture, assumptions, and failure modes in the current project.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-15
- Stars
- ★ 2
- Forks
- 1
01 前言 Front Matter
- name
- challenge
- description
- Ask Kimi to challenge the architecture, assumptions, and failure modes in the current project.
- allowed-tools
- mcp__plugin_kimi-relay_relay__start_task, mcp__plugin_kimi-relay_relay__get_task
- argument-hint
- [specific decision or risk area]
- disable-model-invocation
- true
03 SKILL.md 正文
Start a challenge task using the project directory already configured on the plugin MCP server. Use $ARGUMENTS as the focus and background: false, so the call waits and hands back the finished critique in one step; it can take several minutes, and the client backgrounds it on its own rather than blocking the session. Pass background: true only when the user wants to come back to it later. When presenting results, distinguish verified defects from speculative risks.
Pass model or thinkingEffort only when the user named one themselves. Never infer either from the repository or from what seems appropriate — the point of a second opinion is that you did not choose how it was produced.
04 同仓库更多 Skills
status
Show status for a Kimi task or list recent Kimi tasks.
setup
Check whether Claude Kimi Relay, Git, Node.js, and Kimi Code are ready.
review
Ask Kimi to perform an independent read-oriented code review of the current project.
result
Retrieve and evaluate the final result of a Kimi task.
delegate
Delegate an implementation task to Kimi inside an isolated repository copy and return a reviewable patch.
cancel
Cancel a running or queued Kimi task.