Guide product managers through PM workflows using the PM Brain framework library. Use when working on product management tasks, braindumping ideas, assessing opportunities, writing PRDs, conducting research, or planning strategy. Supports thinking-first approach before jumping to templates.
Politics & organization survival: use the politics-coach skill with 01-Company-Context/1.1-Stakeholder-Avatars/ (person-level avatars) and 01-Company-Context/1.2-Organization-Survival/ (system-level politics: power map, alliances, red flags) when stakeholder problems are clearly political, not just about message shape.
Typical File Patterns
Most framework folders follow this structure:
1-*-framework.md - Guide explaining the framework
2-*-template.md - Template to fill out
3-*-evaluation.md - Assessment criteria
Braindumping Workflow
When a user wants to work on something or is thinking/braindumping (the agent is in product_sense): apply the golden rule from PRODUCT-SENSE-RULES.md (braindump before structure), including the "braindump sufficient" checklist. Use prompts from 02-Methods-and-Tools/2.0-Foundations/2.0.1-Mental-Models/6-Product-Sense-Development/2-product-sense-prompts.md for the relevant context (PRD, prioritization, strategy, research, stuck).
What stage are they at? (ideation, validation, building, shipping)
What constraints exist?
Guide exploration (do not suggest templates yet)
Use prompts from 2-product-sense-prompts.md to surface assumptions and blind spots
Ask clarifying questions from the relevant framework
Help organize scattered thoughts only after raw thinking is out
Suggest framework (only after braindump / when leaving product_sense into execution_mode)
Match their need to the right framework location
Show the framework guide first (1-*-framework.md)
Only then point to the template (2-*-template.md)
Modes & evals
product_sense → braindump, prompts, no framework until sufficient. execution_mode → structure + framework/template (or template-finder path). meta_reflection → after substantial conversations: suggest 00-Meta/ (log, forecast, learning), optionally Level 2 checklist (.cursor/evals/), and rule updates (.cursor/rules/thinking.mdc). Full routing and states: ORCHESTRATION.md.
Signal mode transitions in natural language when switching (per AGENTS.md and ORCHESTRATION.md); e.g. "We've got enough to structure this—here's the framework that fits…"
Evals are a separate workflow, not a conversation mode: Level 1 (artifact quality) lives in 02-Methods-and-Tools/ and the agent uses Quick Quality Checks per .cursor/rules/evaluation-orchestration.mdc when creating supported frameworks; Level 2 (agent behavior) lives in .cursor/evals/. See ORCHESTRATION.md → Eval Checkpoints. You may suggest the Level 2 checklist in meta_reflection; the user runs evals when they choose.
Common Scenarios
"I have an idea"
→ Start with Discovery (use discovery-research skill):
Problem-Solution Space to separate problem from solution
Before touching templates, ask 2–3 lightweight preflight questions (why this/why now, know vs guess, who it’s for) and a context/memory check, for example:
"Why this, why now?"
"What do you already know vs what are you guessing?"
"Who is this primarily for to read and approve?"
"Do you want to anchor this in any existing strategy/initiative/research, or keep this PRD self-contained for now?"
PRD templates and guides
User Stories for requirements
Metrics for success criteria
"How do I convince stakeholders?"
→ Look in Communication (use stakeholder-management skill; add politics-coach when dynamics are clearly political):
One-Pagers for executive summaries
Stakeholder Management strategies
Saying No frameworks for managing requests
"I'm stuck / not sure how to think about this"
→ Start with Foundations:
Mental Models for frameworks
Bias awareness for blind spots
Self-Reflection for clarity
"I'm overwhelmed with requests"
→ Treat this as an overwhelm / paralysis case:
First, acknowledge the feeling and keep cognitive load low.
Ask at most 1–2 gentle questions to narrow, for example: "What’s one thing that, if you made a bit of progress on it this week, would make you feel less stuck?"
Help the user choose a tiny, concrete next step (e.g. "open the first onboarding email and jot 3 bullets on what feels off") instead of introducing mini‑frameworks.
Make explicit that they choose: "You choose the smallest step that feels doable; I’ll help you shape it."
"Am I improving? / How do I track my judgment?"
→ Point to 00-Meta:
Product Judgment Test: 00-Meta/0.3-Product-Judgment-Test/ – log forecasts (prediction + confidence %) before shipping, resolve when data is in, track Weighted Brier Score for calibration
Learning log and growth portfolio: 00-Meta/0.1-Learning-Log/, 00-Meta/0.2-Growth-Portfolio/