对大量原始聊天记录进行结构化整理,进行知识提取或生成任何主题分析文档
npx skills add https://github.com/cafe3310/public-agent-skills --skill long-chat-task-processor
此技能旨在将非结构化的、按标题组织的聊天记录(Markdown格式)转化为可执行的项目管理资产。它严格基于文档目录结构 (TOC) 进行分段处理,而非简单的行数切分,以确保对话上下文的完整性。
当用户提供导出的聊天记录(Markdown),且记录使用标题(#, ##...)区分不同群聊或对话对象时。
用户通常要求:
首先,必须初始化工作区并解析文档结构。
运行脚本扫描源文件标题结构,并生成工作区:
python3 .gemini/skills/long-chat-task-processor/scripts/setup_workspace.py "path/to/chat_log.md" "工作区目录名称"
工作区目录名称可使用 YYYY-MM-DD-HH 沟通记录整理 格式。
初始化后,工作区结构如下:
Chat_Projectization_YYYY-MM-DD-HH-MM/
├── 0-工作日志.md # [核心] 基于 TOC 生成的树状任务列表
├── 1-原始记录/ # 存放源文件
├── 2-项目背景/ # 存放用户提供的背景文档,以及用户的所有指示
├── 3-实体映射表.md # [动态] 自动积累的人名/概念术语表
├── 4-任务池.md # [动态] 累积提取的任务列表
├── 5-决策与里程碑.md # [动态] 累积提取的决策和时间点
└── 6-额外输出/ # 存放用户要求的额外的输出文档
2-项目背景/ 目录下,确保后续处理有据可依。打开 0-工作日志.md,你将看到一个基于 Markdown 标题层级的任务树。
按顺序处理每一个标记为 [ ] 的 Section。
在处理每个 Section 前,务必读取:
0-工作日志.md (获取当前 Section 的行号范围、标题背景)2-项目背景/ (理解业务上下文)3-实体映射表.md (确保人名对齐)处理步骤:
Line Start-End,读取 1-原始记录/ 中对应的内容。4-任务池.md。格式:[ ] <Time> Assigner -> Assignee: <Task> (Status)5-决策与里程碑.md。3-实体映射表.md。6-最终输出/ 下创建或追加对应的文档(例如 6-最终输出/API_Issue_Log.md)。0-工作日志.md 中将该 Section 标记为 [x]。当所有 Section 处理完毕后:
4-任务池.md,合并重复项,按人名或优先级归类。4、5 和 6 中的内容进行汇总。-- 日期 或 -- 日期 时间 (如 -- 02-09 15:00) 来标记时间点,可用于参考。Status: UNCONFIRMED。2-项目背景/ 和 3-实体映射表.md 以恢复上下文。然后继续处理 0-工作日志.md 中未完成的 Section。关于 0-工作日志 ,
可以类似这种格式,在 Agent 处理完成后,将对应的 [ ] 改为 [x] 。
- [ ] **群聊:API 稳定性治理** (Line 100-167)
- [ ] **私聊:小张** (Line 168-600)
关于 3-实体映射表 ,
可以包含人和人的角色、群、组织、项目、概念等的解释,作为多次 Agent 工作之间的上下文补充。
不要使用表格。
关于 4-任务池,
应该包含
事项可以按如下格式记录:
<时间> 交代人 -> 接收人 <事项>
- 来源群或单聊:XXX
- 交付物
- 是否确认和认可
- 后续状态
Prepare meeting materials with Notion context and Codex research; use when gathering context, drafting agendas/pre-reads, and tailoring materials to attendees.
> This skill generates, creates, or builds brand voice guidelines from source materials. It should be used when the user asks to "generate brand guidelines", "create a style guide", "extract brand voice", "create guidelines from calls", "consolidate brand materials", "analyze my sales calls for brand voice", "build a brand playbook from documents", "synthesize a voice and tone guide", or uploads brand documents, transcripts, or meeting recordings for brand analysis. Also triggers when the user has a discovery report and wants to convert it into actionable guidelines.
Prepares meeting materials by gathering context from Notion, enriching with Claude research, and creating both an internal pre-read and external agenda saved to Notion. Helps you arrive prepared with comprehensive background and structured meeting docs.
Share a Google Docs document with edit access and email collaborators the link.
> Lightweight NDA, MSA, and vendor contract review for SMBs without legal on staff. Reads contracts from local files, Gmail attachments, or DocuSign envelopes; flags non-standard terms; explains risks in plain English; and outputs a marked-up redline as a separate DOCX. Use when the user says "review this contract," "what am I signing," "red flags," "flag any concerns," "check the payment terms," or uploads/forwards a contract or legal agreement.
Search across all connected sources in one query. Trigger with "find that doc about...", "what did we decide on...", "where was the conversation about...", or when looking for a decision, document, or discussion that could live in chat, email, cloud storage, or a project tracker.
Check the status of existing agreements with a vendor across all connected systems — CLM, CRM, email, and document storage — with gap analysis and upcoming deadlines. Use when onboarding or renewing a vendor, when you need a consolidated view of what's signed and what's missing (MSA, DPA, SOW), or when checking for approaching expirations and surviving obligations.
Ingest supplied source material into an Obsidian vault with provenance and claim tracking: pasted text, files staged in the selected vault's inbox or .raw archive, or explicitly approved URLs. Use for a single source or bounded batch, not for saving an assistant answer. Triggers: ingest, ingest this file, ingest this URL, process this source, read and file this source, batch ingest, ingest these sources.
Take cafe3310/long-chat-task-processor from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.