メニュー
One paragraph. State precisely when the agent should activate this skill. Include trigger phrases and contexts. Be specific — vague descriptions prevent auto-discovery.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Extract text, tables, citations, forms, and metadata from PDFs as grounded context for downstream M365 Copilot work. Use whenever a source for research-synthesis, decision-memo, or document-summarization arrives as a PDF (contracts, regulatory filings, policy documents, vendor reports, scanned forms, fillable intake forms, academic papers), including encrypted, permission-restricted, hybrid native-plus-scan, or large multi-hundred-page packs. Triggers: "extract from this PDF", "summarize this filing", "pull table from PDF", "OCR this scan", "read the fillable form", "this PDF is password-protected".
Produce a recap of decisions, actions, and owners — not a transcript. Use within 24h of any meeting where decisions were made or actions assigned. For a recurring series, also produce a private behavioral readout over a set window — speaking-time, decision velocity, and action carryover as a share of the whole, each finding tied to one concrete change. Triggers: "recap the meeting", "send notes", "what did we decide", post-meeting summary, "are our standups working", recurring-meeting health, meeting-series readout.
Critically inspect Copilot's draft for facts, tone, omissions, bias, and audience fit, then triage what you find by confidence and severity. Use as the REVIEW step of the pipeline, before any iteration or shipping. Triggers: before refining; before send; before a decision or sign-off; after every Copilot draft of meaningful length; reviewing a deliverable, deck, memo, or generated code/automation; running specialized passes (callout accuracy, silent gaps, will-it-survive-scrutiny, structure integrity).
Meta-skill for creating and improving dojo skills. Use when you've found a workflow worth packaging — typically after a pattern recurs 3+ times in tasks/lessons.md — or when an existing skill triggers unreliably and needs a test-driven tune-up. Triggers: "should this be a skill", recurring pattern, promoting a lesson, packaging a workflow, "this skill never fires", optimizing a skill description for discovery.
Apply your organization's brand - colors, typography, logo, and tone - consistently across Copilot-generated deliverables in Word, PowerPoint, Teams, and Viva. Use when polishing a deck, document, or post so it looks on-brand and accessible. Triggers: "make it on-brand", "apply our branding", "brand colors", "company template", "style this deck", "corporate identity", "match our style guide".
Phased workflow for building a high-quality connector or MCP server that lets Copilot act on an external system (an internal API, a line-of-business app, a SaaS tool). Use when planning, designing, or reviewing custom connector tools for Copilot or M365 agents. Triggers: "build a connector", "MCP server", "expose our API to Copilot", "custom tool for Copilot", "connect Copilot to <system>", "Graph connector design".
| name | <skill-name-in-kebab-case> |
| category | <kata | waza | kumite-task | kumite-office | kumite-connector | cowork | do> |
| description | One paragraph. State precisely when the agent should activate this skill. Include trigger phrases and contexts. Be specific — vague descriptions prevent auto-discovery. |
One-line statement of what this skill does and why it matters.
