12 Claude Skills for Futures Work That Changes a Decision

What these skills are

Signals of change into scenarios, scenarios into moves someone owns. It starts the same way every time. A client, or your own partners, want "something about the future" — a trends piece, a future-of session, a slot in the offsite — and nobody has said what decision it should change. So someone opens a deck, writes five trends everyone already believes, and the room nods at them for forty minutes and goes back to work unchanged.

One line holds the whole pack together: real, dated signals are not optional, and no scenario is ever a forecast.

So the pack starts before the trends. focal-question-framer turns the appetite into one focal question with a horizon, a scope, and the decision it must inform — and it hands back three candidate questions, not one, so the sponsor has something to argue with. signal-scanner then researches the web and logs concrete, dated, sourced signals, one card per signal. A signal it cannot link goes into a "needs a source" list and is never counted as evidence.

From there the work climbs. trend-synthesizer clusters the cards into named trends with an honest evidence grade, and files anything thin as a hunch rather than promoting it. uncertainty-mapper separates the settled from the genuinely open and proposes the two axes the scenarios will stand on. scenario-builder writes three or four constructed futures a leadership team can stand inside — same detail, same plausibility, no ranking — and future-artifact-maker makes them tangible enough to react to.

Mechanically, each skill is one folder with a SKILL.md file. One frames the question, three gather and sort the evidence, two build the futures, three test the strategy and turn it into moves with owners and dates, and three take the whole thing to a room — the trend report, the workshop and the leadership playback. They run from one Claude Project per engagement, with a pinned chat per artifact, and each reads what the ones before it wrote, by file name — so the signal log keeps growing, the scenarios get revised rather than rebuilt, and the signposts have a real history by the second quarter.

Download all 12 skills. One zip: ready-to-install skill zips, readable SKILL.md files, and the two templates the pack runs on — the engagement brief and the blank signal log. Free, no signup. The same 12 skills are open on GitHub: polar-bear-org/claude-skills.

The 12 skills you get

One frames the question, three gather and sort the evidence, two build the futures, three test the strategy and turn it into moves, and three take it to a room.

1 · Frame the question

1. Focal Question Framer

Use when: A client or your own partners want "a trends piece" or "a strategy offsite about the future" and nobody has said what decision it should change

Output: One focal question with a time horizon, a scope, and the decision it must inform — written into the engagement brief every other skill reads, with three candidate questions rather than one so the team has something to argue with

2 · Gather the evidence

2. Signal Scanner

Use when: You have a brief and need evidence, or your team keeps saying "everyone knows that" without a single link

Output: A signal log: concrete, dated, sourced signals of change, one card per signal, across social, technological, economic, environmental and political territory, graded weak, emerging or established — plus a "needs a source" list and a counter-signals section

3. Trend Synthesizer

Use when: You have forty signal cards and need to know what they add up to before the workshop

Output: Named trends with a direction, a maturity and an honest evidence grade. A trend needs three signals from two independent sources; anything below that is filed as a hunch, kept and named, not promoted

4. Uncertainty Mapper

Use when: The trends are on the wall and the team is about to argue about which ones matter

Output: What is already settled separated from what could go either way, the drivers sorted by impact and uncertainty on a three-point scale, the two axes your scenarios will stand on — and a log of who argued for what

3 · Build the futures

5. Scenario Builder

Use when: The axes are agreed and you need worlds people can walk into, not bullet lists

Output: Three or four internally consistent constructed futures, each with its logic, its timeline, who wins and who loses, and the first signs you would see — every one labeled a constructed future, built to the same level of detail, never ranked by likelihood

6. Future Artifact Maker

Use when: The scenarios read as reasonable and nobody feels anything yet

Output: Artifacts from inside each world — a headline, a product listing, a complaint letter, an annual report excerpt, a job ad — clearly marked fictional, in formats the room already reads every week

4 · Test the strategy, and act on it

7. Strategy Wind Tunneler

Use when: Leadership asks "so are we fine?" and you want a better answer than "probably"

Output: Your current strategy, or a set of options, run through every scenario and graded holds, conditional or fragile, with the trigger that would force a change — and a guard against the team's favorite answer

8. Implications Roadmap Builder

Use when: The offsite is over, the energy is high, and Monday is coming

Output: Moves, not themes: no-regret moves now, options to keep open, and big bets with triggers — each with a verb, an owner, a start date, a review date and the scenario evidence it rests on. It does not choose the bet for you

9. Signpost Tracker

Use when: Three months after the workshop, someone asks "which future are we in now?"

Output: Early-warning indicators for each scenario with thresholds and a review cadence, plus the quarterly re-scan and review that keeps the scenarios and the roadmap alive instead of archived

5 · Take it to the room

10. Trend Report Writer

Use when: Marketing wants "our annual trends report" and you want it to be worth reading

Output: The publishable report your firm sends to clients: each trend framed as a tension with a question for leaders, its sourced signals, the implications and a provocation — and a refusal to print a trend with no sourced signals behind it

11. Scenario Workshop Runner

Use when: You are running the session yourself and want it to produce arguments, not sticky notes

Output: The half-day or full-day design: agenda, materials, facilitation prompts and capture, for a room that sorts the uncertainties, builds or tests the scenarios and claims the first moves. The room itself stays human

12. Leadership Playback Designer

Use when: The work is done and the only thing left is the meeting that decides whether any of it mattered

Output: The pre-read, the 45-minute arc, the one decision you are asking for, and the capture afterwards — with the favored world never presented last, and no probability attached to any scenario, even when the CEO asks

How an engagement runs through the pack

One Creative Futurology Studio project holds the engagement, named after the client or the firm and the year. The filled engagement brief sits in its knowledge — the focal question, the horizon, what is in and out of scope, the decision this work must inform and who owns it, who is in the room and who is missing, and what the firm already believes — and every skill reads it before it starts. Then one pinned chat per unit of work, named after the artifact it produces.

The first chat runs focal-question-framer, which will tell you plainly whether you are ready to scan or still looking for the decision this work should change. Scanning comes next, one round per chat, and it is where the pack is strictest: no source, no signal. Trends are built on top of the log, and a cluster with fewer than three signals from two independent sources is filed as a hunch — kept, named, and barred from becoming an axis or a chapter until the evidence catches up.

Then the futures. uncertainty-mapper proposes the axes and records the argument instead of replacing it; the team moves the cards and the moves are logged with who argued for them. scenario-builder writes the worlds, each labeled a constructed future, each at the same level of detail, none ranked by likelihood. strategy-wind-tunneler runs the current plan through every one of them and grades it holds, conditional or fragile, with the trigger that would force a change. implications-roadmap-builder turns that into no-regret moves, options and big bets with owners and dates — and stops short of choosing the bet. signpost-tracker sets the thresholds and comes back every quarter.

It stops where your judgement starts. The pack does not forecast numbers: no market sizing, no growth projections, no econometric models. It does not run customer or user research. It does not execute the roadmap — once the moves have owners and dates they belong to your normal management rhythm. And it does not predict: ask any of the twelve which future will happen, and you get the same answer, kindly, plus what it can do instead.

Setup guide

1. Download the pack. One zip: an install folder with 12 ready-to-upload skill zips, a skills folder with the same skills as readable files, and the two templates the pack runs on — engagement-brief.md and signal-log.md.
2. Install your skills. In Claude Code, add the marketplace and install creative-futurology-pack, and all 12 load at once. In Claude, turn on code execution in Settings, Capabilities so the skills can save their artifacts as files, 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 Creative Futurology Studio. Make one Claude Project per engagement and name it "Creative Futurology Studio: [Client or Firm] [Year]". Upload the filled engagement brief to the project's knowledge, and add each artifact as it is produced, so the later skills can read the earlier work.
4. Open a chat per unit of work, and frame the question. Pin one chat per artifact — "brief", "signals round 1", "trends", "uncertainties", "scenarios", "artifacts", "wind tunnel", "roadmap", "signposts Q1", "report", "workshop", "playback" — and start in the first one with "run focal-question-framer". Each skill saves its output as [artifact]-[slug].md (signals-acme-2031.md, scenarios-acme-2031.md), and every later skill reads the earlier ones by those names. Every skill also runs on its own: if an artifact is missing it asks for the few things it needs, carries on, and says it worked without it.

Where to start

The quality bar

Every skill in the pack holds the same standard, the one we hold when we do this work ourselves:

• No source, no signal: a signal has a date and a link a colleague can open, and anything unlinkable goes to a separate "needs a source" list instead of being counted as evidence
• Nothing is logged from memory: with web search off, the scanner says so at the top of its output and works only from the documents you give it
• A trend needs three signals from two independent sources; below that it stays a hunch — named and kept, but never used as a scenario axis or a chapter of the report
• What a colleague heard is kept and labeled hearsay, and never counted toward a trend's evidence threshold until a source is attached
• Every scenario is labeled a constructed future, built to the same level of detail as the others, and never ranked by likelihood — no scenario gets a probability, even when the CEO asks for one
• Future artifacts are marked fictional before they are made, and what must never appear — a real client, a real person — is agreed up front
• Drivers are scored low, medium or high rather than to two decimals, because false precision is worse than none, and a team's placement is never overwritten with the skill's own
• The argument is recorded rather than replaced: who argued for a placement, and what the argument was, so the reasoning survives the workshop
• A move is something a person can begin on a Tuesday: a verb, an owner, a start date, a review date, and a cost in a unit the firm already uses — never a number the skill invented
• Winners, losers and capability gaps are types of firms and roles, never named individuals, and nothing produced in the work is used to assess a participant

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 turns an appetite for the future into a decision someone owns. When you want your whole team working this way, AI carrying the scanning and the drafting so people do the arguing and the choosing, across strategy, delivery, and everyday work, that's what we build with clients.

Meet Pauline. Staring down a futures brief nobody has framed yet? Bring the offsite, the trends piece, or the decision it is supposed to change. Book a 30-minute call · Pauline on LinkedIn