xz1220

oh-my-kimi 的目录入口，包含面向 Kimi CLI 的 agent、skill、hook 与 MCP 套件，衍生自 oh-my-* 谱系。

面向密钥、注入、authz/authn、不安全 IO、依赖与数据外泄风险的安全评审

证据驱动的追踪通道，在 Kimi 的 Agent 工具中编排互相竞争的 tracer 假设

LLM Wiki —— 跨会话持续累积的 markdown 知识库（Karpathy 模型）

面向作家的 agentic 记忆系统 —— 跟踪人物、关系、场景与主题

跑只读的深度仓库分析，返回一份带置信度排序的综合结论，附具体文件引用、清晰区分证据与推断。当用户说 'analyze'、'investigate'、'why does'、'what's causing'，或在提任何改动方案之前需要跨文件的有据解释时使用。

取消任何活跃的 oh-my-kimi 模式（autopilot、ralph、ultrawork、ecomode、ultraqa、swarm、ultrapilot、pipeline、team）

2 阶段 pipeline：trace（因果调查）→ deep-interview（需求结晶），含 3-point injection

Write and run dynamic workflows: short JavaScript scripts in Claude Code's workflow dialect, run with the `odw` CLI outside the host agent's context, fanning subtasks out to coding-agent CLIs (Codex, Claude Code, Gemini, Qwen, Kimi, or your own) in a background process and handing back only the final result. Use this skill when a task is bigger than a single call — parallel draft fan-out, multi-stage review pipelines, adversarial verification of findings, or discovery loops that run until nothing new turns up — or when the user mentions odw, dynamic workflows, multi-agent orchestration, or fanning out subagents.

编写并运行 dynamic workflow：用 Claude Code 的 workflow 方言写一段简短的 JavaScript 脚本，再用 `odw` CLI 在宿主 agent 的上下文之外，把子任务扇出给 coding-agent CLI （Codex、Claude Code、Gemini、Qwen、Kimi 或自定义），后台跑完后只取回最终结果。 当一个任务大过单次调用——需要并行扇出多份草稿、多阶段评审流水线、对抗式核验发现， 或循环挖掘直到无新发现——或用户提到 odw、dynamic workflow、多 agent 编排、扇出 subagent 时，使用本 skill。

Execute explicitly authorized Ultracode / Ultra Mode requests: optimize for exhaustive, correct answers within the user's stated scope by using any available independent-work mechanism (ODW workflows, subagents, dispatched workers, or tool-worker fan-out), with serial multi-pass fallback only when no such mechanism exists. Use when the user explicitly says ultracode, ultra mode, go exhaustive, deep mode, multi-agent, parallel, adversarial, comprehensive, do not miss anything, check every angle, or asks for several independent verification strategies. Do not activate merely because a task is difficult, technical, ambiguous, or high-stakes.

执行被显式授权的 Ultracode / Ultra Mode 请求：在用户给定范围内，把优化目标设为尽可能 详尽和正确，使用当前环境提供的任何独立工作机制（ODW workflow、子智能体、派发 worker、 并行工具任务）进行覆盖、验证和综合；只有完全没有这类机制时才使用串行多趟退化方案。 当用户明确说 ultracode、ultra mode、go exhaustive、深度模式、多 agent、并行、 对抗式、彻底处理、别漏掉任何东西、每个角度都查一遍，或要求多种独立验证策略时使用。 不要只因为任务困难、技术性强、含糊或高风险就自行启用。

Run a read-only Kempt health check on the local AI-coding-agent state (~/.claude, ~/.codex) and explain the agent-debt report. Use when the user asks why their coding agent feels slow, confused or inconsistent, mentions "agent debt" / "agent 体检" / "agent health", or wants to know what junk has piled up in their Claude Code / Codex config.

Reversibly clean the agent debt that Kempt found — stale sessions, unused skills, stale/duplicate memories, orphan caches — with everything moved to a recycle bin that supports one-command undo. Use when the user asks to clean up their Claude Code / Codex state, free agent context budget, or act on a Kempt report ("清理 agent 债", "clean my agent", "apply the kempt plan").

Task routing — Claude tokens are expensive, Codex usage is cheap. When an "execution-style" task comes in (bulk file edits, translation, long-form/sample code/test generation, format conversion, data cleanup, image generation, large-output reconnaissance, web research), read this skill first and decide whether it can be offloaded to codex exec. Claude only breaks the task down, writes the work order, and handles verification. If Claude gets stuck twice, it can also ask Codex for a second opinion on stuck tasks. Use this when the user says "use codex", "save tokens", "offload to codex", or /codex-offload. The following tasks should not be offloaded and should be handled by Claude directly as usual (this is not a ban): tasks that need session context or memory, architecture decisions, outward-facing actions, and small jobs that can be finished in a few minutes.

任务分流：Claude token 贵、Codex 用量便宜。接到"执行型"任务（批量改文件、翻译、长文/样板代码/测试生成、格式转换、数据清洗、生成图片、大输出侦查、联网调研）时，先读本 skill 判断能否外包给 codex exec，Claude 只负责拆解、写工单和验收；Claude 卡壳 2 次的难题也可丢给 Codex 换视角。用户说"用 codex"、"省 token"、"外包给 codex"或 /codex-offload 时也使用。以下任务不外包、由 Claude 照常亲自执行（不是禁令）：需要会话上下文或记忆的任务、架构决策、对外发送类操作、几分钟就能干完的小活。

AI-powered side hustle discovery. Build your profile, get personalized recommendations with quantified scoring. | AI 副业发现引擎：建立画像、个性化推荐、量化评分。

用 lark-switch 选定某个飞书/Lark 账号（多租户）来执行 lark-cli 命令。当任务涉及多个飞书账号、需要以某个指定账号执行 lark-cli 或 lark-* skill、用户提到 lark-switch / 切换账号 / 多账号 / 别的租户身份，或需要确认当前命令实际跑在哪个账号上时使用。不负责 lark-cli 本身的登录授权排障（走 lark-shared）。

写飞书文档（技术方案、PRD、复盘、说明等）时遵守的写作规范，目标是让人读一遍就懂。当用户要写、改、审阅或润色一篇飞书 / 技术文档，问"这么写清不清楚""怎么把流程讲明白"，或显式输入 /feishu-doc-style 时使用。核心：只留结论不写推理、段落两三行、流程用图不用文字、不用黑话、格式克制、标题写内容不写结构。

溯源镜 / TraceLens — 给一个消息链接或一段内容,做联网深度溯源:提炼核心论断、全网检索、按时间线还原"谁最早发的、谁在转",判定有没有可信一手来源、检测"语义漂移",最后产出一份可本地保存的精美 HTML 报告(不在终端堆分析)。Use when the user pastes a news/social link or a claim/message/小作文 and asks to 溯源 / 核查 / 验证真伪 / 查出处 / 看是不是假消息 / 看怎么传出来的, or to trace a rumor, fact-check a story, or find its original source.

Use the Agent Xiangqi CLI to join and play real-time Xiangqi matches as an Agent.

Use the Agent Poker CLI to join a Texas Hold'em match, wait for live turn updates, reason about legal actions, and submit the next action.

Give the agent cross-session emotional memory and proactive, non-intrusive care for the user. Activate whenever the user shares anything personal — emotions, health, relationships, or upcoming/past significant events — or when starting a new session where prior personal context exists at ~/.claude/personal-context/. Records to a user-global store so care continues across projects.

Convert media links or local audio files into Chinese transcripts using the local link-transcript skill workflow and FunASR SenseVoiceSmall. Use when the user provides a Douyin/TikTok/Bilibili/social media URL, MP3, MP4, or audio/video file and asks to extract speech, transcribe, convert to text, or get the spoken content.