Agent Skills

---

name: cavekit-methodology

description: Cavekit specification-driven development methodology — the Hunt lifecycle (Draft → Architect → Build → Inspect → Monitor) and how to apply it. Use when starting a Cavekit project, structuring an existing codebase as kits/plans, or routing between sub-skills.

---

name: cavekit-revision

description: Trace bugs and manual fixes back to kits and prompts; fix at the source so the iteration loop can reproduce the fix autonomously. Six-step revision process plus the single-failure backpropagation protocol. Use when a manual hot-fix has been applied, when convergence stalls, or when the same class of bug keeps reappearing.

---

name: cavekit-validation-first

description: Validation-first design for Cavekit — every kit requirement must be automatically verifiable. Six-gate validation pipeline, phase gates, merge protocol, completion signals, acceptance criteria patterns. Use when designing acceptance criteria, wiring CI gates, or debugging why an agent is producing output that nobody can verify.

---

name: caveman-compress

description: Ultra-compressed communication mode (lite / full / ultra) that cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Use when the user requests "caveman mode", "less tokens", "be brief", or when output budget is tight.

---

name: forgecad

description: ForgeCAD model authoring, editing, debugging, and execution guidance for .forge.js, SVG-import, assembly, and CLI workflows. Use when building or modifying ForgeCAD geometry, structuring multi-file projects, validating scripts, or using ForgeCAD export/render tooling.

---

name: forgecad-build-model

description: Build or edit a manufacture-realistic `.forge.js` model in a project, then validate it with run, render, inspect, and export evidence.

---

name: forgecad-design-spec

description: Create a ForgeCAD design brief, HLD, or LLD before coding by walking through use, assembly, interfaces, decisions, and verification.

---

name: forgecad-grade-model

description: Grade a ForgeCAD or CAD-as-code model against a requirement, brief, prompt, reference, or acceptance criteria with evidence and a 0-10 score.

---

name: forgecad-image-prompt

description: Write builder-honest AI image prompts from a concrete ForgeCAD model, build brief, HLD, or LLD without hiding how the artifact is built.

---

name: forgecad-inspect-model

description: Select, run, and interpret ForgeCAD inspection evidence for collisions, sections, wall thickness, components, masks, depth, normals, surface continuity, and fit.

---

name: forgecad-project-sync

---

name: forgecad-reconstruct-cad-file

description: Reconstruct a readable parametric ForgeCAD model from an existing CAD or mesh file such as STL, OBJ, 3MF, STEP, or STP.

---

name: forgecad-reconstruct-from-images

description: Reconstruct a real parametric ForgeCAD object from reference images by using images as evidence, not as a one-view facade.

---

name: forgecad-verify-mujoco

description: Verify a ForgeCAD MJCF export in MuJoCo with dynamics, contacts, controls, joint travel, and rendered evidence before calling it simulation-ready.

---

name: hf-mem

description: Hugging Face CLI to estimate the required memory to load Safetensors or GGUF model weights for inference from the Hugging Face Hub

---

name: plugin-creator

description: Scaffold a complete cave plugin bundle — generates .cave-plugin/plugin.json manifest and the standard directory structure (commands/, skills/, agents/, themes/, hooks/). Use when a user wants to create, publish, or package a cave plugin for the marketplace. Triggered by "create a plugin", "scaffold a plugin", "/plugin create", or "new cave plugin".

---

name: ak-rss-digest

description: Curate a Chinese reading digest from a fixed bundle of RSS and Atom feeds, with a strong preference for AI agent thinking, frontier AI commentary, deep interviews, and non-boring high-signal essays. Use when Codex needs to pull the latest week's posts by default, or a specific day's posts when explicitly requested, summarize them, score each article on a 10-point scale, and output only the posts scoring above 7 in a concise Chinese daily-brief style.

---

name: anthropic-style-diagram

description: Draw architecture, flow and structural diagrams in the Anthropic/Claude visual language as SVG, then render and save them as PNG. Trigger ONLY when the user names that visual language — 'Anthropic 风格架构图', 'Claude 风格流程图', 'Anthropic style diagram', 'Claude style flowchart', 'Anthropic 配色的图' — or asks for another diagram in the same style after this skill has already produced one in the conversation. Also use to re-render or export an SVG that already uses this skill's class system (th/ts/box/arr/c-blue…). Do NOT use for a plain 画图/画流程图/draw a diagram request with no style named, for data charts, UI mockups, illustrations, TikZ or mermaid output, or database ERDs.

---

name: daily-news-report

description: 基于预设 URL 列表抓取内容，筛选高质量技术信息并生成每日 Markdown 报告。

---

name: gemini-watermark-remover

description: Remove the visible Gemini AI watermark from images using reverse alpha blending. Use when asked to strip Gemini watermarks, batch-process Gemini images, or build/modify a CLI script that removes the bottom-right Gemini watermark without HTML or server-side components.

---

name: transcript-polisher

description: 将语音转录文本（访谈、演讲、播客、会议）精修为可读性更高的文章段落。当用户提到"字幕精修"、"transcript polish"、"润色字幕"、"把视频字幕整理成文章"、"访谈文字整理"、处理访谈记录、转录文本优化、语音转文字整理、或者需要将大段对话/演讲文本整理成可读文章时触发。适用于单人演说或多人对谈的转录文本整理，要求保留原句原词、拒绝高度概括。即使用户只是说"帮我整理一下这段文字"并附上了明显的口语化文本，也应该触发此技能。

---

name: translate-polisher

description: 高质量文章翻译技能，采用"分析→初译→审校→终稿"四步精翻工作流。仅支持中文↔英文、中文↔日文翻译。当用户明确提出"翻译"、"translate"、"精翻"、"翻訳"、"翻译文章"、"translate to Chinese/English/Japanese"、"改成中文"、"改成英文"、"改成日文"、"翻成中文"、"翻成日文"、"翻成英文"、"英译中"、"中译英"、"中译日"、"日译中"、"日本語に翻訳"、"中国語に翻訳"、"英語に翻訳"、"これを翻訳して"、"put this in Chinese"、"put this in English"、"put this in Japanese"、"convert to Chinese"、"convert to English"、"convert to Japanese"、"帮我翻一下"、"本地化"、"localize"、"这篇文章翻译一下"，或给出 URL/文件/正文并明确要求输出目标语言成稿时触发。不用于仅做摘要、解释、理解或整理的请求。若输入是 URL，优先使用 `curl -L` 请求 `r.jina.ai` 抓取正文 Markdown；抓取失败或正文不完整时必须直接停止并要求用户自行提供正文。

---

name: web-to-markdown

description: Convert a web URL into cleaned Markdown with deterministic routing. Use when Codex needs to read article-like content from links and should apply source-aware fetch strategies: default to r.jina.ai for general pages (including X/Twitter), use defuddle.md for YouTube links, and use browser-impersonated extraction for WeChat/Zhihu/Feishu pages with Mozilla Readability cleanup.

---

name: clawhip

description: Attach/install/use clawhip as a daemon-first notification gateway runtime for OpenClaw / Clawdbot