Skills

Twelve PM skills that encode expert frameworks, failure modes, and quality gates. When an agent knows how an expert thinks about a problem, it stops producing generic output and starts producing defensible analysis.

12 skills · ~1,200 lines each · shared/toolkits/skills/

The problem

AI gives you generic output because it lacks domain expertise. Ask it to do competitive analysis and you get a Wikipedia summary. Ask it to write a spec and you get vague requirements.

Skills encode how an expert thinks about a problem — frameworks, failure modes, quality gates. Not what to produce, but how to reason. Load a skill and the agent stops guessing at methodology and starts applying the right one.

What gets encoded

Twelve PM skills across four categories, each a standalone markdown file with ~1,200 lines of encoded methodology. Extensible to any domain you know deeply.

Analysis & Intelligence

Competitive Market Analysis

9 frameworks: 7 Powers, Aggregation Theory, Wardley Mapping, JTBD, Blue Ocean, Crossing the Chasm, and more. Produces a Competitive War Map with evidence tier annotations.

Discovery Research

Evidence triangulation, interview analysis, research quality assessment, signal vs noise classification. Turns raw data into defensible insight.

Problem Framing

Problem Definition Canvas, 5 Whys, opportunity sizing, constraint mapping, ICE/RICE prioritization. Separates the real problem from the presenting symptom.

Strategy & Planning

Product Strategy

Vision-to-roadmap translation, portfolio prioritization, strategic bets framework, platform vs product decisions. Connects business goals to product decisions with evidence.

Go-to-Market Strategy

Launch sequencing, channel strategy, adoption flywheel design, market timing analysis. Builds GTM plans that survive contact with real users.

Pricing & Packaging

Value metric identification, tier design, willingness-to-pay analysis, competitive pricing, packaging for expansion. Turns pricing from guesswork into structured methodology.

Measurement & Specification

Metric Design & Experimentation

North Star metric rubrics, Goodhart's Law countermeasures, A/B test design, retention cohort methodology. Builds measurement systems that don't get gamed.

Specification Writing

Outcome-first methodology, acceptance criteria taxonomy, scope boundary protocol. Produces zero-question documents — specs engineers can build from without a follow-up meeting.

Stakeholder Alignment

Stakeholder mapping, influence strategy, alignment rituals, escalation frameworks, decision rights clarity. Turns political complexity into structured navigation.

Communication & Distribution

Narrative Building

Narrative arc construction, positioning, Why Now analysis, audience adaptation, objection anticipation. Builds arguments that hold up under pressure.

Executive Writing

Board memo structure, executive summary methodology, decision document formats, brevity-with-depth techniques. Writes for people who have 5 minutes and need to make a $10M decision.

Multi-Channel Publishing

Long-form to short-form compression, platform-native adaptation, distribution sequencing, content atomization. One thesis becomes 8 format-native pieces.

Real example

Here's an excerpt from the competitive analysis skill — the structure that routes to the right framework and catches the most common failure modes:

skills/competitive-market-analysis/SKILL.md

## Step 0: Framework Selection
Route to the right frameworks based on the question:
- Market entry?      → 7 Powers + Wardley + JTBD
- Competitor move?   → COAP + Aggregation Theory
- Build/buy/partner? → Blue Ocean + Crossing the Chasm

## Step 1: Context Gate
Before applying ANY framework, verify:
- Is competitive analysis the right artifact?
- Is this user-choice or IT-provisioned?
- Do you have internal data access?

## Failure Modes
FM-1: Framework without evidence (applying 7 Powers
      with no market data)
FM-8: Right frameworks, wrong question
FM-9: Expert-only document (no reader navigation)

The Context Gate exists because the most expensive mistake isn't a bad framework — it's the right framework applied to the wrong question. Agent Prime caught this on a real M365 Copilot analysis: the war map was technically correct but the real problem was internal activation, not competitive positioning.

How to set it up

Each skill is a standalone markdown file that any agent can load. The structure is consistent across all twelve:

Context Gate        → Is this the right artifact?
Framework Selection → Which frameworks apply to this question?
Step-by-step method → Exactly how to apply each framework
Output format       → What a complete output looks like
Failure modes       → What breaks this analysis
Quality gates       → How to know if the output is good

Start by encoding one domain you know deeply. Write down how you actually think when you're doing it well — the sequence of questions, the frameworks you reach for, the mistakes you've seen people make. That's a skill file.

Agent Prime's skills went from 44% baseline to 93% benchmark score. The methodology is the difference, not the model.

What you get

shared/toolkits/skills/

Twelve PM skills across analysis, strategy, measurement, and communication. Each a complete methodology file, portable across agents, extensible to any domain.

12 skills · ~1,200 lines each · ~14,000 lines total