foundation-stakeholder-briefings
について
このスキルは、単一の信頼できる情報源となるドキュメントを作成し、そこから経営陣やエンジニアリング部門など、異なる対象者向けに特化した複数の関係者別ブリーフィングを生成します。派生したすべてのブリーフィングが追跡可能な投影であり、バージョン間の一貫性が保たれることを保証します。同じ核心的な情報を、それぞれ異なる視点や詳細度を必要とする複数のグループに伝える必要がある場合にご利用ください。
クイックインストール
Claude Code
推奨npx skills add product-on-purpose/pm-skills -a claude-code/plugin add https://github.com/product-on-purpose/pm-skillsgit clone https://github.com/product-on-purpose/pm-skills.git ~/.claude/skills/foundation-stakeholder-briefingsこのコマンドをClaude Codeにコピー&ペーストしてスキルをインストールします
ドキュメント
Stakeholder Briefings
A stakeholder-briefings artifact takes one source (a PRD, a discovery synthesis, a research report, a GTM or launch plan, experiment results, a retro or incident write-up, or raw notes) and produces a single saveable file containing:
- a master document: the canonical, audience-neutral synthesis of the work (what and why, decisions, status, risks and open questions, asks, timeline), with each claim numbered (
M1,M2, ...); and - a set of audience briefings: the same content re-pitched for each chosen stakeholder, one self-contained, copy-paste-ready block per audience.
The skill runs master-first, then projects. The master is the single source of truth; every briefing is a projection of it. A briefing may omit, reorder, and translate master content, but it may never assert a claim that is not in the master. That projection rule is what keeps the executive version and the engineering version from quietly disagreeing, and it is the difference between this skill and asking a model to "rewrite this six ways."
Distinct from foundation-stakeholder-update (one async update of meeting outcomes for a single audience, meeting-bound), discover-stakeholder-summary (a map of who stakeholders are and their influence/interest), and foundation-persona (a customer/buyer viewpoint to design or market against).
When to Use
- One piece of work must reach several audiences who each need a different framing, decision, and level of detail (a spec going to engineering, design, data, and the funder at once).
- You are about to manually rewrite the same update three to five ways, one per audience.
- A decision or result needs to propagate across functions without the versions drifting apart.
- A single audience needs a tailored briefing from a non-meeting source (N=1 is supported; the fan-out is the signature use, not a floor).
When NOT to Use
- One async update of meeting outcomes for stakeholders. Use
foundation-stakeholder-update(it is meeting-bound; that is its scope). - Understanding or mapping stakeholders (influence, interest, comms plan). Use
discover-stakeholder-summary. - A persona to design or market against. Use
foundation-persona. - There is no source content yet. This skill projects an existing artifact; it does not do the underlying analysis.
Instructions
When asked to create stakeholder briefings, follow these steps:
-
Ingest and classify the source. Read the provided artifact. Classify its type (spec/PRD, discovery/research, GTM/launch, strategy/roadmap, experiment/metrics, incident/retro, compliance/privacy/security, or raw/ambiguous). If the source is thin, continue but set
input_quality: lowand name the gap. -
Build the master. Write the audience-neutral canonical document with these sections: What and Why, Decisions, Status, Risks and Open Questions, Asks, Timeline. Number every load-bearing claim with a stable ID (
M1,M2, ...). The master carries no audience-specific spin; it is the shared substrate. -
Propose the audiences. From the source type, propose the relevant subset using
references/source-type-map.md(for example, a spec proposes Engineering, UX/Design, Data/BI, Executive; a GTM plan proposes PMM, Sales, CS/Support, Executive). Present the proposal and acceptgo(generate the proposed set), an edit (drop X, add Y), orall(all nine). If invoked with--go, skip the prompt and generate the proposal. No audience is ever locked out. -
Project each briefing. For each chosen lens (see
references/audience-lenses.md), render a self-contained block delimited by--- BEGIN: <lens> ---/--- END ---, containing:Draws on:the master claim IDs this briefing projects (required).Primary ask:exactly one decision or action for this audience (required).- a one-line headline, a "what this means for you" framing, and the body, at the lens's length, vocabulary, and tone. Every load-bearing line must trace to a master claim. Do not introduce a claim that is not in the master.
-
Flag translations. Keep a translations-applied log (internal, below the shareable boundary) for every technical-to-business or inferred re-pitch, so the user can verify it lands. This section is never part of a shareable briefing.
-
Self-check the invariant before finalizing:
- Trace references resolve (deterministic, checkable): every briefing
Draws on:ID resolves to a real master claim. - One CTA (deterministic, checkable): exactly one
Primary ask:per block. - No untraced claim (review): re-read each block against its
Draws on:set and confirm the body introduces nothing absent from those master claims. This is a review step, not automated. - Neutral master (review): the master has no audience-specific spin. List anything that fails.
- Trace references resolve (deterministic, checkable): every briefing
-
Render the artifact. Master (with claim IDs) -> the delimited briefing blocks -> the boundary marker -> the translations-applied log -> Sources and References. Remove all guidance blockquotes from the final output.
Audience lenses
Nine first-class lenses, each defined by the decision it owns, plus a Custom slot whose lens is inferred from the audience name and source and shown for confirmation. Full definitions, per-lens "not this lens when" boundaries, and the overlap matrix (Exec vs Board, PMM vs Sales, Engineering vs Data, Legal vs Exec) are in references/audience-lenses.md.
Output
- A single artifact (filename
YYYY-MM-DD_HH-MMtz_<title>_stakeholder-briefings.md), built fromreferences/TEMPLATE.md. - Each briefing block is self-contained and send-ready (BEGIN/END cut-lines) so it can be copied out without edits.
- A future
--splitmode (write each block to its own file) is deferred; v1 is single-artifact.
Quality checklist
- Master present with numbered claim IDs (
M1,M2, ...) and no audience-specific spin. - Each briefing block has a
Draws on:line whose IDs all resolve to master claims. - Each briefing block has exactly one
Primary ask:. - No briefing asserts a claim absent from the master (projection rule).
- Audience set matches the source-type proposal or the user's edit; N=1 honored without refusal.
- Translations-applied log present (internal) when any translation was made; boundary marker separates shareable blocks from internal sections.
- Each briefing is at the lens's length and tone (a board block reads nothing like an engineering block).
- Guidance blockquotes removed from the final artifact.
See also
references/TEMPLATE.md- the master + briefing-block scaffold.references/audience-lenses.md- the nine lenses, boundaries, and overlap matrix.references/source-type-map.md- the source-type to audience proposal.foundation-stakeholder-update- one meeting update for one audience (distinct).discover-stakeholder-summary- mapping stakeholders (distinct).
GitHub リポジトリ
Frequently asked questions
What is the foundation-stakeholder-briefings skill?
foundation-stakeholder-briefings is a Claude Skill by product-on-purpose. Skills package instructions and resources that Claude loads on demand, so Claude can perform foundation-stakeholder-briefings-related tasks without extra prompting.
How do I install foundation-stakeholder-briefings?
Use the install commands on this page: add foundation-stakeholder-briefings to Claude Code as a plugin, or clone its repository into your skills directory, then restart Claude so it picks up the skill.
What category does foundation-stakeholder-briefings belong to?
foundation-stakeholder-briefings is in the Design category, tagged word, ai and data.
Is foundation-stakeholder-briefings free to use?
Yes. foundation-stakeholder-briefings is listed on AIMCP and free to install. It runs inside Claude, so no separate service account is required to use the skill itself.
関連スキル
executing-plansスキルは、完全な実装計画があり、それを管理されたバッチでレビューチェックポイントを設けながら実行する場合に使用します。このスキルは計画を読み込んで批判的にレビューした後、小さなバッチ(デフォルトは3タスク)でタスクを実行し、各バッチの間に進捗状況を報告してアーキテクトのレビューを受けます。これにより、品質管理チェックポイントが組み込まれた体系的な実装が保証されます。
このスキルは、コードレビュアーサブエージェントを起動し、処理を進める前に要件に対してコード変更を分析します。タスク完了後、主要な機能の実装後、またはmainブランチへのマージ前などに使用すべきです。このレビューは、現在の実装と元の計画を比較することで、問題を早期に発見するのに役立ちます。
このスキルは、開発者がHTTP、stdio、またはSSEトランスポートを使用してMCPサーバーをClaude Codeに接続するための包括的なガイドを提供します。GitHub、Notion、カスタムAPIなどの外部サービスを統合するためのインストール、設定、認証、セキュリティについて解説しています。MCP統合のセットアップ、外部ツールの設定、またはClaudeのModel Context Protocolを扱う際にご利用ください。
このスキルは、タスク分析に基づいて開発者がClaude Code WebとCLIインターフェースの選択を支援し、これらの環境間でのシームレスなセッションテレポーテーションを可能にします。Web、CLI、モバイル環境を切り替える際のセッション状態とコンテキストを管理することで、ワークフローを最適化します。様々な段階で異なるツールを必要とする複雑なプロジェクトにご活用ください。
