charpup
- 6 skills
- 0 followers
- 1 day ago last updated
- ▌ Clean Web Markdown · charpup bundleRetrieve clean markdown from webpages for AI agents using protocol-level content negotiation and crawl-to-markdown fallbacks. Use when users ask to fetch/read/summarize websites with low-noise markdown, to improve web_fetch quality, or to process JS-heavy pages via Jina/Firecrawl. Chinese trigger cues include: "抓网页正文", "提取网页 Markdown", "网页转 Markdown", "读取网页并总结", "这个链接帮我清洗一下".
- ▌ Inbox Triage · charpup bundleTRIGGER: inbox-triage, 查收inbox, inbox分拣, inbox triage, 处理inbox. 扫描 inbox/ 目录，AI 分类每个文件，展示分拣计划，询问是否创建新项目目录， 执行文件移动并验证结果。适用于定期清理 inbox 的完整分拣流程。
- ▌ Task Workflow · charpup bundleTRIGGER: task scheduling, dependency analysis, DAG scheduling, batch ordering, task dependencies, complexity scoring, multi-task project. Intelligent task scheduling with DAG-based dependency resolution, complexity scoring, and batch ordering. Reads tasks from task_plan.md (planning-with-files output), builds a dependency graph, and produces an optimal execution schedule. Use when managing 5+ interdependent tasks. NOT for simple sequential task lists.
- ▌ Skill Review · charpup bundleSkill 生命周期审计与部署的完整流程编排。当 inbox/ 目录出现新的 SKILL.md 文件、 用户想评估一个 skill 是否值得安装、需要判断某个 skill 的安全性或工作区适配性时， 必须触发此 skill。自动编排以下完整流程： (1) 用 Explore 子代理读取 skill 文件， (2) 调用 skill-vetter 安全审计， (3) 执行四维分析（设计意图/代码质量/可执行性/适配性）， (4) 调用 value-first-gate 价值评估（含正反驳斥）， (5) GO/NO-GO 部署决策， (6) 安装到 ~/.claude/skills/ 或归档，并写入 value-review.md， (7) 调用 inbox-triage 清理 inbox。 触发词：评估这个 skill、这个 skill 值得装吗、审计 skill、vet this skill、 review this skill、inbox 有新 skill、help me evaluate a skill、skill 安全吗。 不要等用户手动编排每个步骤——只要有 skill 评估需求就主动触发完整流程。
- ▌ Summarize · charpup bundleRecord session deliverables (files created/modified and cross-session decisions) to a long-term memory file. Use when the conversation has produced file outputs or decisions with cross-session impact. Skip for pure Q&A conversations. Trigger on "/summarize" command or: 保存成果、归档本次对话、存入记忆、 记录这次的产出、写入长期记忆、保存决策。
- ▌