菜单
Use this skill whenever the user wants to write, edit, format, summarize, comment on, or review the Word document open in the side pane. Routes user intent to the single right office_* tool so the model picks one high-leverage call instead of chaining many low-level reads and writes.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/Agents365-ai/innate-work --skill edit-document命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Create new pi skills from a conversation. Use whenever the user says they want to make a skill, build a skill, create a skill, save a workflow as a skill, or asks "how do I make a skill that does X" — even if they don't say the word "skill" but describe a recurring task they want to capture once and reuse. Make sure to use this skill whenever the user mentions creating, authoring, drafting, or building a skill in any phrasing.
Extract clean markdown content from web pages using Defuddle CLI, removing clutter and navigation to save tokens. Use when the user provides a URL to read or analyze, for online documentation, articles, blog posts, or any standard web page. Do NOT use for URLs ending in .md — those are already markdown, fetch them directly with your normal web-fetch tool.
Use when the user asks Innate Desktop to work with their connected Obsidian vault — "save this as a note", "add to my Obsidian", "find my notes on X", "what did I write about Y". Here the vault is NOT the working directory; it lives at the absolute path given in your system prompt. Read and write notes (Markdown files) under that path, preserving Obsidian conventions.
基于 SOC 职业分类
正在显示 SKILL.md
| name | edit-document |
| description | Use this skill whenever the user wants to write, edit, format, summarize, comment on, or review the Word document open in the side pane. Routes user intent to the single right office_* tool so the model picks one high-leverage call instead of chaining many low-level reads and writes. |
| license | MIT |
| metadata | {"category":"office-routing","primary-interface":"office_* tools"} |
| triggers | ["Word","document","文档","Word文档","写一份","修改","总结","评审","排版","套模板"] |
This document is open in Word right now. You have office_* tools that read and write it directly. Route the user's intent to one well-chosen tool before reaching for many.
| User intent | First tool | Notes |
|---|---|---|
| "Write me a report / proposal / outline / draft / 写一份" (no current content matters) | office_build_paragraphs | Single call. Pass the full structure as an array of paragraphs each with its own style and list mode. Do NOT call office_get_document_outline or office_get_document_text first — the document is empty or will be replaced. |
| "Rewrite this paragraph / make this concise" (selection-driven) | office_replace_selection | Read selection with office_get_selection only if you need surrounding context. |
| "Change all X to Y" (global text edit) | office_find_replace | One call. Don't pre-read the full text. |
| "Edit paragraph N / the second heading" | office_replace_paragraph_by_index | Call office_get_document_outline first to map text to paragraph index. |
| "Add a comment / review note / 批注" | office_add_comment | For "comment on this paragraph / the selected text", OMIT anchor so the whole selection is covered. Pass anchor only to target a different specific phrase (first match, that substring only). |
| "Reply to / resolve comments" | office_get_comments → office_reply_comment / office_resolve_comment | |
| "Summarize / explain / 总结 / 解释" | office_get_document_outline first, then office_get_document_text only for the sections the user asked about | Never load full text just to summarize a large doc. |
| "Make this bold / italic" | office_apply_inline_format | |
| "Make this Heading 2 / Quote" | office_apply_paragraph_style | |
| "Turn these lines into a bullet / numbered list" | office_apply_list | Or pass `list: "bullet" |
| "Accept / reject 修订 / tracked changes" | office_get_revisions then office_resolve_revision({ index, action: "accept"|"reject" }) | For "全部接受 / accept all changes", pass action: "accept-all" AND confirmAcceptAll: true only after the user explicitly asks for it. Requires WordApiDesktop 1.1+. |
| "Add / insert a link / 加链接 / hyperlink" | office_insert_hyperlink | Pass anchor to wrap an existing phrase; omit to wrap the current selection. displayText overrides what the link shows. |
| "Insert image / 插入图片" | office_insert_image | Bytes must already be base64-encoded — pi-office does NOT fetch URLs. Always include altText unless the image is purely decorative. |
| "Make a new table / 做个表 / N × M table" | office_build_table | Single call with a 2D array. Use office_set_cell only to update an existing table. |
"诊断 / debug / 排查 / 为什么连不上 / /pioffice:doctor" | office_doctor | Read-only. Prints a multi-line health report; show the full result back to the user verbatim so they can act on the ✗ rows. |
Reflexive read before a write. If the user said "write me a 5-section product brief", go straight to office_build_paragraphs. Reading the document first wastes a turn.
Chaining N small writes instead of one batch. office_build_paragraphs accepts paragraphs with per-item style and list mode. Use it for any multi-paragraph composition. Do not insert one paragraph, then apply a style, then insert another paragraph.
Re-pasting the result in chat after a write. After office_build_paragraphs, the user sees the document. Reply with at most ~3 lines:
[Section](pi://goto?heading=Section),Asking for confirmation before reading. Read tools (office_get_*) are silent and safe. Use them freely; do not ask "should I read the document first?"
Calling office_get_document_text on a long doc. Use office_get_document_outline first. Only fetch text for the section the user asked about.
office_build_paragraphs({position: "replace-document"}) wipes the entire document body before inserting. The tool refuses this call unless you also pass confirmReplaceDocument: true. Only set that flag when the user has clearly said something like "重写整篇 / 替换文档 / rebuild from scratch / start over". If they only said "draft / write / add", use position: "end" (default) or position: "start" instead — those preserve existing content.
office_get_document_outline returning [] means "blank document, no need to preserve anything." Skip straight to office_build_paragraphs.
If the request has a genuine fork you cannot settle from context — two or more plausible directions that would change what you build — call office_ask_user once (pass options for a choice, multiSelect to let them pick several). Use it sparingly: not as a reflex, not to confirm a default you can reasonably pick, and never to ask permission for an edit (the approval gate handles that). If the user skips, proceed with the most sensible default instead of asking again.
Pick the single tool whose name most directly matches the user's verb:
office_build_paragraphsoffice_replace_*office_find_replaceoffice_add_comment / office_get_commentsoffice_get_revisions → office_resolve_revisionoffice_insert_hyperlinkoffice_insert_imageoffice_build_table (new) or office_set_cell (update)One well-chosen tool beats five guesses.