AgentSkillsCN

strategy-doc

运用Cagan + Playing to Win框架,结合人机协作,制定产品战略文档。

SKILL.md
--- frontmatter
name: strategy-doc
description: Develops product strategy documents using Cagan + Playing to Win frameworks with human-in-the-loop collaboration

Strategy Doc Writer

Build rigorous product strategy documents through structured research, collaborative development, and critical review.

When to Use

User wants to:

  • Develop a product strategy
  • Create a strategic planning document
  • Think through a major product decision
  • Document vision, positioning, or competitive strategy

Frameworks

This skill combines three proven frameworks:

Cagan (Product Strategy)

  • Vision: 2-5 year north star
  • Principles: Decision guardrails
  • Strategic bets: Sequenced problems to solve

Playing to Win (Competitive Strategy)

  • Where to Play: Target customers, markets, channels
  • How to Win: Competitive advantage, what makes us hard to copy
  • Capabilities: What we must build/acquire

Amazon FAQ

  • Anticipated objections
  • "What if we're wrong about X?"

Final Document Structure

code
1. Vision & Principles (Cagan)
2. Where to Play (P2W)
3. How to Win (P2W)
4. Strategic Bets (Cagan)
5. Capabilities Required (P2W)
6. FAQ (Amazon)

Workflow

CRITICAL: Follow this workflow exactly. Each phase has a human gate. Save artifacts as you go.

PHASE 0: SETUP

code
□ Create working directory: scratch/strategy-[topic]-[date]/
□ Initialize this checklist in the working directory

Actions:

  1. Ask user: "What strategic question are we tackling?"
  2. Create working directory with slugified topic name
  3. Save initial checklist to progress.md

PHASE 1: FRAME (fast)

code
□ Capture the strategic question in one sentence
□ User confirms we're pointed at the right problem
□ Save: 00-frame.md

Actions:

  1. Restate the strategic question in one clear sentence
  2. Ask user to confirm or refine
  3. Save frame to working directory

Human gate: User confirms framing before proceeding


PHASE 2: GATHER (critical step)

code
□ Fetch relevant Notion context
□ Research competitors and market landscape
□ Pull relevant metrics/data
□ Synthesize into research brief
□ Save: 01-research-brief.md
□ User reviews research, flags gaps or redirects scope

Actions:

  1. Spawn agents IN PARALLEL using Task tool:
    • notion-researcher: Find existing strategy docs, past decisions, related context
    • competitor-researcher: Market landscape, competitor positioning, trends
    • Use analyze skill if user has relevant data questions
  2. Synthesize agent outputs into a research brief
  3. Save research brief to working directory
  4. Present brief to user for review

Human gate: User reviews research, flags gaps, may redirect scope

Agents to spawn:

  • .claude/agents/notion-researcher.md
  • .claude/agents/competitor-researcher.md
  • (Phase 6) .claude/agents/notion-writer.md

Notion researcher output format for this skill:

markdown
## Existing Context: [Topic]

### Relevant Prior Decisions
- **[Decision]:** [Brief context and date]

### Active Constraints
- **[Constraint]:** [Why it matters]

### Relevant Prior Research
- **[Topic]:** [Key findings, source doc]

### Open Questions from Prior Work
- **[Question]:** [Context]

### Source Documents
| Document | Last Updated | Link |
|----------|--------------|------|

PHASE 3: DEVELOP

code
□ Draft Vision statement options
□ Define Principles (decision guardrails)
□ Explore Where to Play options
□ Define How to Win for chosen arena
□ Sequence Strategic Bets
□ Identify Capabilities required
□ Save: 02-strategic-choices.md
□ User approves strategic choices

Actions: Work through each element interactively:

  1. Vision: Propose 2-3 vision statement options. Ask user to pick or refine.
  2. Principles: Suggest 3-5 decision guardrails based on research. User confirms.
  3. Where to Play: Present options for target customers/markets/channels. Discuss trade-offs.
  4. How to Win: For chosen arena, articulate competitive advantage. Pressure-test: "What would have to be true?"
  5. Strategic Bets: Sequence the problems to solve. Why this order?
  6. Capabilities: What must we build or acquire to execute?

Save choices to working directory after user approval.

Human gate: User approves strategic choices before drafting


PHASE 4: DRAFT

code
□ Write full document with FAQ
□ Save: 03-draft-v1.md
□ User reviews draft

Actions:

  1. Spawn strategy-writer agent with:
    • Research brief (01-research-brief.md)
    • Strategic choices (02-strategic-choices.md)
    • Document structure and writing style guidelines
  2. Save draft to working directory
  3. Present draft to user

Human gate: User reviews draft, provides feedback

Agent to spawn:

  • .claude/agents/strategy-writer.md

PHASE 5: REVIEW

code
□ Strategy-reviewer critiques document
□ Save: 04-review-feedback.md
□ User incorporates feedback
□ Save: 03-draft-v2.md (iterate as needed)

Actions:

  1. Spawn strategy-reviewer agent to critique the draft
  2. Save review feedback to working directory
  3. Present feedback to user
  4. User decides which feedback to incorporate
  5. Revise draft (spawn strategy-writer again if major changes)
  6. Save revised draft with incremented version

Human gate: User satisfied with draft before publishing

Agent to spawn:

  • .claude/agents/strategy-reviewer.md

PHASE 6: PUBLISH

code
□ Confirm stakeholders and Notion location
□ User approves final
□ Convert to Notion format and publish
□ Save final: final.md

Actions:

  1. Ask user: "Who are the stakeholders for this doc?"
  2. Ask user: "Where in Notion should this live?" (get page ID or parent page)
  3. Get explicit approval to publish
  4. Spawn notion-writer agent to:
    • Verify formatting is Notion-compatible
    • Publish to specified Notion location
    • Return page URL
  5. Save final version to working directory
  6. Confirm publish succeeded with link

Human gate: Explicit approval before any Notion write

Agent to spawn:

  • .claude/agents/notion-writer.md

Critical Rules

  1. NEVER skip human gates. Each phase requires user confirmation before proceeding.
  2. ALWAYS save artifacts. Every phase produces a file in the working directory.
  3. Spawn agents in parallel when possible (Phase 2 research agents).
  4. Read the working directory if resuming a session to understand current state.
  5. Use the checklist in progress.md to track what's done.

Resuming a Session

If user returns to continue a strategy doc:

  1. Ask which strategy doc they're working on (or check scratch/ for recent directories)
  2. Read progress.md to see current phase
  3. Read latest artifacts to restore context
  4. Continue from where they left off