12 Claude Skills for an Innovation Program That Earns Its Budget

What these skills are

A program that can say what each bet cost and what it proved. Most firms of 20 to 200 people do not have an innovation portfolio; they have a pile. Six things are "in progress", four of them are the same kind of bet, one is a founder's long-held dream that has never been tested, and one is an internal tool that quietly became a product without anyone deciding it should. Nothing gets killed, because killing something takes a rule somebody wrote down while calm.

One position holds the pack together: evidence comes from the market, and decisions come from named humans.

Claude designs the test and keeps the ledger. It never simulates a customer, scores an idea, or makes the kill call. The interviews happen with real people, the experiments cost real money, and every gate decision carries the name of the person who made it. What the pack carries is the paperwork that usually decides whether a program survives its second year: the charter, the gate rules, the experiment cards, the decision records and the ledger.

Mechanically, each skill is one folder with a SKILL.md file. Two write the program's rules once, three run on a cadence — the ledger monthly, the portfolio map quarterly, the investment case quarterly and yearly — and seven run per initiative, in order, from the challenge to the sunset. Each reads the artifacts the ones before it wrote, from one Claude Project per program, so the charter, the cards and the signed decisions stay in one findable place.

Download all 12 skills. One zip: ready-to-install skill zips and readable SKILL.md files. No templates to fill in — each skill reads what the earlier ones wrote, by file name. Free, no signup. The same 12 skills are open on GitHub: polar-bear-org/claude-skills.

The 12 skills you get

Two skills write the program's rules, two keep it honest month to month, five carry one bet from a challenge to evidence, and three turn that evidence into a decision, a budget, or a dignified close.

1 · Write the rules, once

1. Program Charter Writer

Use when: You keep starting side projects and nobody can say what the program is for or who owns the money

Output: The one page that lets a program say no: why it exists, what counts as an initiative, the target ambition mix, the money and time envelope, who decides what, and the review rhythm. Run it first — every other skill in the pack reads the charter

2. Stage Gate Designer

Use when: Your gate meeting either waves everything through or kills the interesting ones

Output: Three or four stages with one decision per gate, questions instead of scores, a small funding round per stage, a way back one stage instead of dying, and kill criteria agreed while nobody's project was on the table

2 · Keep the program honest

3. Innovation Ledger Keeper

Use when: Leadership asks what the program delivered and all you have is a list of activities

Output: The program's accounting: leading indicators per initiative and per stage, plus program-level measures — cycle time, kill rate, learning velocity, money by stage. One line per initiative per month, the same columns every time. First run after the charter and the gates, then monthly

4. Portfolio Mapper

Use when: You have nine things going and suspect they are all the same kind of bet

Output: Every live initiative placed on ambition and stage, the mix held against what the charter promised, and the crowding and the holes named in plain words. It does not rank initiatives; it shows the firm what shape its bets have

3 · Take one bet from challenge to evidence

5. Innovation Challenge Framer

Use when: The partners want you to "do something with AI for clients" and that is the whole brief

Output: A problem statement with edges: who has the problem, what it costs them, what a good answer would have to be true of, and what is out of bounds. Deliberately narrow — narrow challenges produce testable ideas, wide ones produce posters

6. Ideation Session Runner

Use when: You have a challenge, a Friday afternoon and twelve people

Output: The format the challenge needs — pitch hour, sketch sprint, worst-idea round, client-signal harvest — with selection rules that keep the partner's opinion from being the vote, and an exit that is two or three idea one-pagers rather than a wall of notes. Claude does not generate the ideas

7. Discovery Interview Planner

Use when: Somebody should talk to customers before the prototype, and nobody has said who or what to ask

Output: The conversations an idea needs before anything is built: who, how many, the guide, consent, the note template, and the synthesis into evidence. Claude plans them and turns the notes into evidence; it never has the conversation or plays the customer

8. Experiment Designer

Use when: You think agencies would pay for this, and you want to find out in three weeks without building it

Output: The assumptions mapped, the riskiest one picked, and an experiment card naming the method, the number that means pass, the number that means fail, and the date — all written before the test runs, because afterwards every number looks like the one you meant

9. Experiment Results Reader

Use when: The landing page got 41 signups and everyone in the room has an opinion about whether that is good

Output: The results held against the card the team wrote first, with neither side renegotiating the numbers: the learning in one paragraph, and a scale, pivot, persevere or kill recommendation for a named human to decide on. Results with no prior card are a story, and get treated as one

4 · Decide, fund, or close

10. Gate Review Preparer

Use when: The gate review is Thursday and the team has thirty slides

Output: One initiative ready for one gate: the one-page evidence brief in the shape every initiative uses, the gate's own questions and kill criteria to be read before the team speaks, a ten-minute script, and the decision record with one name on it

11. Investment Case Writer

Use when: Budget season, and the program has to justify its existence in a way that survives the CFO

Output: The case for continued investment in the currency the program actually produces, which is reduced uncertainty: what was learned, what was stopped and why, the portfolio against the target mix, and what the next round costs — in the units the charter used. It will not dress a quarter with no decisions as momentum

12. Initiative Sunset Writer

Use when: You are stopping the client-portal idea and the two people who ran it are taking it personally

Output: The memo that records the decision and the honest reasons, the harvest of what can be reused, and the message to the team and to the customers who took part. It writes the decision as the decider's, never as the people's — and it will not run without a signed one

How a bet moves through the program

One Innovation HQ project carries the program and every bet inside it. In the Program chat, program-charter-writer runs first and writes the charter everything else reads. stage-gate-designer follows, once, before the first initiative enters the pipeline. innovation-ledger-keeper sets the columns after both and then adds a line a month, and portfolio-mapper runs quarterly — and before any challenge round, so the next challenge aims at a hole instead of a crowd.

Each initiative gets its own pinned chat, opened with the challenge. innovation-challenge-framer writes it; ideation-session-runner refuses to run a session without one, and turns the afternoon into an idea one-pager per idea worth testing. discovery-interview-planner plans the conversations before anything more expensive than a conversation gets built, and experiment-designer writes the card for the riskiest assumption still standing.

experiment-results-reader opens that card when the results arrive and holds the two next to each other. gate-review-preparer reads the gate rules for the question, the results chain for the evidence and the ledger for spend and cycle time, then writes the one-page brief before the review and the decision record, signed, after it. Nothing in the brief comes from anywhere but those files and the team's answers.

Two skills close the loop. investment-case-writer runs quarterly for the portfolio review and yearly for the budget, reading the ledger, the map, every decision record and sunset memo in the period, and the previous case — so the new one answers what the last one promised. initiative-sunset-writer runs after a decision to stop or park, and not before one exists, and writes the memo, the harvest and the message to the team.

The pack ends where the scaling decision is made: building, launching and running the thing that passed the last gate is product and delivery work with its own craft. It ships the customer research an innovation program needs — problem interviews and evidence synthesis — and not the wider design research discipline: no usability testing, no ethnography, no personas, which belong with the Design Thinking Pack. It does not cover open innovation with outside partners, corporate venture investing, or the legal and IP work of a spin-out; those need specialist counsel.

Setup guide

1. Download the pack. One zip: an install folder with 12 ready-to-upload skill zips, and a skills folder with the same skills as readable files. There is nothing else to fill in — this pack ships no templates, because each skill reads the artifacts the earlier ones wrote by file name.
2. Install your skills. In Claude Code, add the marketplace and install innovation-pack, and all 12 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. Create your Innovation HQ. Make one Claude Project per innovation program — for most firms, that is one — and call it Innovation HQ. Keep one pinned chat called Program for the charter, the gates, the ledger, the portfolio map and the investment case, and one pinned chat per initiative, named after the initiative's slug, where the challenge, the ideas, the interviews, the experiments, the gate briefs and the sunset live in order. Every skill saves its output as a markdown file named for its artifact (charter-innovation-program.md, experiment-checkout-1.md, decision-checkout-gate2.md), and later skills read the earlier ones by those names. Then open the Program chat and write "run program-charter-writer". Everything else reads the charter.

Where to start

The quality bar

Every skill in the pack holds the same standard, the one we hold when a firm is spending real money to find something out:

• Evidence comes from the market and decisions come from named humans: Claude designs the test and keeps the ledger, and never simulates a customer, scores an idea, or makes the kill call
• Kill criteria are agreed before anything enters the pipeline — a criterion invented during the review is an opinion with a number on it
• One gate asks one question and produces one decision, on a brief shaped the same for every initiative, so reviewers compare evidence and not presentation skill
• Pass and fail thresholds are written on the card before the experiment runs, because afterwards every number looks like the one you meant
• Results with no prior card are a story rather than an experiment, and get read as one
• The ledger is boring by design: one line per initiative per month, the same columns every time — a ledger whose columns change every quarter is a slideshow
• The portfolio map shows the shape of the firm's bets; it never ranks them
• Every gate decision carries a name, and a stop with no signature is not written up, because an initiative stopped without one gets quietly restarted by whoever loved it
• An initiative is closed with the decision separated from the people: how a firm stops things decides whether anyone proposes the next one

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 runs one firm's innovation program. When you want your whole company working this way, AI carrying the overhead so people do the thinking, across innovation, hiring, and everyday work, that's what we build with clients.

Meet Pauline. Nine side projects and nothing killed? Bring the list, the budget you cannot justify, or the gate meeting that decides nothing. Book a 30-minute call · Pauline on LinkedIn