PM Skills Arsenal

AI produces polished analysis that changes no decisions. These skills fix that.

Process guides raise the floor. Knowledge weapons raise the ceiling.

Why This Exists

When you load a skill, the cheapest model beats the best model without it.

Condition Score vs. Baseline
Baseline (Claude, no skill) 47 / 105 —
Anthropic’s PM Skill (methodology) 81 / 105 +72%
PM Skills Arsenal (codex) 98 / 105 +109%

Benchmark: 5 prompts × 7 dimensions × 3 conditions. Full methodology and outputs in benchmark/.

The gap isn’t model quality — it’s encoded knowledge. Raw prompting produces generic analysis. Skill-loaded agents produce structured frameworks with evidence tiers, confidence calibration, and outputs that hold up to scrutiny.

Install (Claude Code) 

claude plugin marketplace add avyayalaya/pm-skills-arsenal
claude plugin install pm-skills@avyayalaya

Skills auto-activate when you ask Claude to do anything in their domain. No slash commands needed.

GitHub Copilot

Reference skill files with # in Copilot Chat or Copilot CLI. Full setup guide: GitHub Copilot Installation

Other AI tools (ChatGPT, Cursor, Gemini): Copy the relevant SKILL.md directly into your context window.

Skills

12 codex skills covering the full PM lifecycle — from analysis through strategy, communication, and influence.

Analysis & Research

Competitive & Market Analysis — Structural assessment of who wins, why, and what to do about it. 7 Powers, Aggregation Theory, Christensen Disruption, JTBD, Wardley Mapping, Blue Ocean, Crossing the Chasm, Win/Loss.

Discovery & Research — Research synthesis that separates what users say from what they do. Evidence Triangulation, Interview Analysis, Signal vs Noise, Demand-Side Analysis, Research Gap Mapping.

Problem Framing — Decomposes vague problem areas into evidence-graded problem statements. Problem Definition Canvas, 5 Whys, JTBD Problem Framing, Opportunity Sizing, Constraint Mapping, ICE/RICE.

Definition & Measurement

Specification Writing — Specs so complete the executor starts building without asking a single clarifying question. Outcome-First Methodology, Acceptance Criteria Taxonomy, Scope Boundary Protocol, Failure Condition Design.

Metric Design & Experimentation — Measurement systems that hold up under pressure. NSM + Decomposition Tree, Leading/Lagging Indicators, Goodhart’s Law, Experiment Design, Retention Cohort Methodology, HEART.

Strategy & Planning

Product Strategy — Strategic product roadmaps with confidence-rated bets, option-value sequencing, and explicit deprioritization. Vision-to-Roadmap Cascade, Strategic Bet-Sizing, NOT-Doing Section, Resource Allocation.

Go-to-Market Strategy — GTM strategy with wedge identification, channel unit economics, and launch gating. Market Entry Thesis, ICP Scoring, Channel Strategy, Competitive Positioning (Dunford), Distribution Mechanics.

Pricing & Packaging — Pricing recommendations with model selection, WTP assessment, and sensitivity analysis. Van Westendorp, Gabor-Granger, Good/Better/Best Architecture, Revenue Impact Modeling, AI/SaaS Patterns.

Communication & Influence

Executive Writing — Executive-ready documents — strategy one-pagers, board memos, decision briefs — with role-calibrated framing. Minto Pyramid / SCR, Format Routing, Audience Calibration, Decision Architecture, Zero-Jargon Compression.

Narrative Building — Strategic narratives that make a product feel inevitable. Positioning (Dunford), Why Now, Audience Adaptation, Evidence-Narrative Integration, Objection Anticipation, Competitive Narrative Analysis.

Multi-Channel Publishing — Channel-specific content derivatives from a source document. 7-Step Compression Protocol, Hook Adaptation, Evidence Density Calibration, Source Fidelity Verification, Spoken Script Derivation.

Stakeholder Alignment — Stakeholder alignment strategy with influence mapping, coalition analysis, and sequenced alignment plan. Power-Interest-Position Matrix, Decision Archaeology, Objection Pre-emption, Alignment Monitoring & Recovery.

See It In Action

Interactive showcases — full skill outputs with tabbed navigation, evidence tiers, and framework analysis.

Analysis & Research

Definition & Measurement

  • AI Chat Interface Spec — A spec for a ChatGPT-style interface complete enough that an engineer starts building without questions
  • M365 Copilot Adoption — How to measure whether a $30/user AI copilot is actually delivering value

Strategy & Planning

Communication & Influence

MIT · Parth Sangani