31 Claude Skills for Deciding What to Build

What these skills are

The whole product job, as installable Claude skills. The pack follows the product job from the inbox to the review after launch. It starts where the week starts, with requests you have to sort and say no to. Then it sets direction — vision, strategy, north star, OKRs, roadmap — goes out to customers with interview guides, synthesis, jobs and assumption tests, writes the work down as a PR/FAQ, a prototype brief, a PRD, a story map and user stories, gets a decision that sticks, prepares sales, support and marketing for launch, and checks what the release actually changed.

The red line: real customers are not optional. Claude drafts the documents, the team talks to users, and a named person makes the call.

Each skill is one named method or one artifact, with the inputs it needs, where it breaks, a fixed output template, a finish line and the next skill to run. Each works from what you paste — this week's requests, five interview transcripts, the usage export, the market data you already have — with no connector to install. Nothing here scores, ranks or profiles people: the skills score requests, options, risks and assumptions, key results are graded and team members are not, and interview material is summarised as needs, never as profiles of individual customers.

Mechanically, each skill is one folder with a SKILL.md file. The 31 are grouped into the seven stages of the product job, and each one gives full value on its own: run one, run a stage, or work through the lot in the order the week arrives.

Download all 31 skills. One zip: ready-to-install skill zips, readable SKILL.md files, and the evidence notes behind every method. Free, no signup. The same 31 skills are open on GitHub: polar-bear-org/claude-skills.

The 31 skills you get

The set follows the product job, in seven stages: prioritise and say no, set direction, discover, define, decide and align, launch, and measure and learn.

1 · Prioritise and say no

1. Feature Request Triage

Use when: Sales promised it, a big client wants it, the CEO forwarded it, all this week

Output: Request log, why-behind-it questions, build / later / no sort, reply drafts

2. Cost of Delay

Use when: Everything is urgent and you need a no that holds

Output: Urgency profiles, CD3 order, "yes, if" options, reply to the asker

3. RICE Prioritization

Use when: 40 requests, room for 5, and the loudest voice keeps winning

Output: Scored backlog, confidence notes, out-of-order exceptions, cut list

4. MoSCoW Prioritization

Use when: The date is fixed and scope has to give

Output: Must / Should / Could / Won't list, Must budget check, Won't-this-time list

5. Kano Model

Use when: You need to know which features customers expect and which delight

Output: Paired survey questions, category table per feature, build-first list

2 · Set direction

6. Product Vision

Use when: The team cannot say in one sentence what it is building toward

Output: Vision statement, vision board (target group, needs, product, business goals)

7. Product Strategy

Use when: Everything is high priority and leaders set the roadmap by preference

Output: Diagnosis, guiding policy, coherent actions, "we will not" list

8. North Star Metric

Use when: The team ships a lot and cannot say which number it is moving

Output: North star, 3 to 5 named inputs with definitions, counter-metrics

9. Product OKRs

Use when: The quarter starts and the goals are a list of features

Output: Quarterly objectives, measurable key results, key-result grading notes

10. Product Roadmap

Use when: The roadmap is yours until it is run from the CEO's inbox

Output: Now-next-later board, confidence notes, what is not on it, executive version

3 · Discover

11. Customer Interview Guide

Use when: You have 30 minutes with a customer and cannot waste it on opinions

Output: Screener lines, story-based questions, probes, consent note, notes template

12. Interview Synthesis

Use when: Five interview transcripts and a PRD due Friday

Output: Affinity clusters, need statements with sources, contradictions, open questions

13. Jobs to Be Done

Use when: A request arrives with no known job behind it

Output: Job statements, circumstances, forces for and against switching, alternatives

14. Opportunity Solution Tree

Use when: Discovery keeps getting squeezed and has to fit between sprints

Output: Outcome, opportunity tree, target opportunity, three solution ideas, assumption tests

15. Assumption Mapping

Use when: The team is about to build on beliefs nobody has checked

Output: Desirability, viability and feasibility assumptions, evidence map, riskiest three

4 · Define

16. Working Backwards PR/FAQ

Use when: A prototype is already built and nobody wrote down why

Output: One-page press release, customer FAQ, internal FAQ

17. Prototype Brief

Use when: Leadership saw the AI demo and thinks it ships next week

Output: What it tests, pass line set in advance, what it does not prove, gap-to-production list

18. PRD Template

Use when: The last spec was read five different ways

Output: Problem, goals, non-goals, requirements, open questions, success measures

19. Story Mapping

Use when: The backlog is 200 tickets and nobody can see the whole product

Output: User activity backbone, walking skeleton, release slices

20. User Stories

Use when: Engineers say the requirements are thin but not what is missing

Output: Stories, Given / When / Then criteria, INVEST check, refinement questions

5 · Decide and align

21. Product Role Charter

Use when: You ship story points, not software, and the week is meetings and tickets

Output: What product owns, what others own, stop-doing list, meeting audit

22. DACI Decision

Use when: Sign-off never comes, or comes and is reversed next week

Output: Driver, approver, contributors, informed, inputs, decision date

23. Decision Log

Use when: Settled calls keep coming back as if nobody decided

Output: Dated records (context, decision, status, consequences), reopening rule

24. Pre-Mortem Analysis

Use when: The launch date was promised before anyone asked what could go wrong

Output: Failure story, reasons list, top risks with owners and signals, go / adjust / stop

6 · Launch

25. Competitive Analysis

Use when: Sales keeps saying "the competitor has it"

Output: Landscape table, win / loss themes, where you differ, what not to copy

26. Value Proposition Canvas

Use when: Marketing, sales and product each describe the product differently

Output: Customer profile, value map, fit notes, one message all teams use

27. Go-to-Market Plan

Use when: The product is ready and sales, support and marketing are not

Output: Launch tier, audiences, positioning line, readiness by team, enablement brief

28. Release Notes

Use when: Support hears about changes from customers

Output: Customer notes by change type, internal version for support, who is affected

7 · Measure and learn

29. Success Metrics

Use when: You celebrated a release and still cannot say whether it worked

Output: HEART table (goals, signals, metrics), baseline, target, guardrail

30. Experiment Design

Use when: Two options, a strong opinion on each side, and enough users to test

Output: Hypothesis, variants, primary metric, guardrails, sample check, decision rule

31. Post-Launch Review

Use when: Two weeks after launch customers still use the old workflow

Output: Intended versus actual, why, keep / change list, keep, fix or remove call

Where to start, and what the pack rests on

There is no compulsory sequence. You bring one real thing — this week's requests, the five transcripts, the prototype leadership saw yesterday, the usage export from a launch — and open the skill that matches what hurts this week. The order in the catalogue is the order the week usually arrives in: prioritisation first, because until there is a no that holds every other stage runs from someone else's inbox, then direction, because direction is what the no is said against.

Several of them chain, and each skill names the next one itself. pm-feature-request-triage sorts the week and hands the build pile to pm-rice-prioritization, which orders it and hands the order to pm-product-roadmap. Direction runs pm-product-vision into pm-product-strategy, strategy into pm-north-star-metric and the north star into pm-okrs. Discovery runs pm-customer-interview-guide into pm-interview-synthesis, synthesis into pm-jobs-to-be-done, pm-opportunity-solution-tree into pm-assumption-mapping and the riskiest assumptions into pm-prototype-brief. Writing it down runs pm-pr-faq into pm-prd, the PRD into pm-story-mapping, the map into pm-user-stories and the stories into pm-pre-mortem before anything is promised. pm-daci-decision ends in pm-decision-log so the call stays decided, pm-go-to-market-plan ends in pm-release-notes, release notes end in pm-success-metrics, metrics open pm-experiment-design when a release needs a controlled test to show it caused the change, and pm-post-launch-review sends what you learned back into pm-opportunity-solution-tree.

The pack is research-informed, not validated. A scan of what product managers and product owners say they struggle with set the order of the skills and the words in each "use when": 771 public items collected between 26 and 27 September 2026, kept as 176 pain statements and grouped into 14 struggles, the largest being "everything is urgent and I cannot say no", "the role is meetings and tickets", "the AI prototype is mistaken for the product", "specs get misread" and "no time or access for discovery". The methods themselves come from their originators and from public research — Intercom on RICE, Kano and colleagues (1984) on attractive and must-be quality, Teresa Torres on opportunity solution trees and story-based interviews, the Nielsen Norman Group on interviews and affinity diagramming, Christensen and colleagues on jobs to be done, Rumelt on the strategy kernel, Amplitude's North Star playbook, Google re:Work on OKRs, the Scrum Guide, Klein on pre-mortems, Rodden and colleagues on HEART, Kohavi and colleagues on controlled experiments, USAID's after-action review guidance, and Black Swan Farming on cost of delay and CD3 — with MoSCoW, now-next-later, the Product Vision Board, assumptions mapping, the Test Card, the Value Proposition Canvas, Amazon's working backwards, story mapping, INVEST, Given / When / Then, DACI, decision records, Keep a Changelog and the Pragmatic Institute framework cited for attribution only. The one shared file in the pack, resources/evidence-and-sources.md, lists them source by source with the finding used and the limit on it. Most of them are practitioner or originator descriptions of a method rather than studies of its effect, and no skill here has been tested in a trial.

Setup guide

1. Download the pack. One zip: an install folder with 31 ready-to-upload skill zips, a skills folder with the same 31 skills as readable SKILL.md files, and the evidence notes — resources/evidence-and-sources.md — which name the source behind every method and where it stops.
2. Install your skills. In Claude Code, add the marketplace and install ai-for-product-management-pack, and all 31 load at once. In Claude, turn on code execution in Settings, Capabilities, then go to Customize, Skills and upload one zip per skill from the install folder. Prefer working from files? Add the SKILL.md files to your Project knowledge instead; it works, just less cleanly.
3. Bring one real thing from the week. No connector, no integration, no admin: every skill works from what you paste — this week's requests from sales, support and the CEO's forward, the five interview transcripts, the usage export from the launch. Start with whatever hurts this week and write "run pm-feature-request-triage". Keep customer names and personal data out: the skills summarise what customers need, they never profile who they are.

Where to start

The quality bar

Every skill in the pack holds the same standard, the one we hold when we run product ourselves:

• One method, applied properly: each skill uses the real mechanics of its method — the scales, the grid, the formula, the order of the pass — not a generic "gather, analyse, recommend"
• It tells you where it breaks: every skill says when not to use it and what fits instead
• Same shape every time: a fixed output template and a short "done when" check before handover
• No invented facts: no made-up quotes, customers, numbers or findings — missing data becomes a bracketed placeholder or an open question
• Work is scored, people are not: the skills score requests, options, risks and assumptions, never team members or individual users, and customer material is summarised as needs rather than profiled
• Every output ends with the decision a named person makes, and by when: nothing sends a reply, a no or a launch message on its own
• The red line: real customers are not optional — Claude drafts the documents, the team talks to users, and a named person makes the call

Who made this

Polar Bear is a people ops consultancy for human-size teams (20 to 200 people). Built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. This pack is the free, self-serve version of how we work.

The pack carries the product job. When you want your whole team working this way, AI carrying the overhead so people do the part only people can do, across product, people and everyday operations, that's what we build with clients.

Meet Pauline. Everything urgent, the roadmap run from someone else's inbox? Bring this week's requests, the strategy you cannot say no against, or last month's launch. Book a 30-minute call · Pauline on LinkedIn

Install in Claude Code

Two lines, and every skill in the pack loads at once.

/plugin marketplace add polar-bear-org/claude-skills
/plugin install ai-for-product-management-pack@polar-bear-skills

Using Claude on the web instead? Download the zip and upload each skill from its install folder under Customize → Skills.