What these skills are
The work a business graduate is hired to do, as installable Claude skills.
This is Claude for the work a business graduate is actually handed: a company to understand before a call, an export with a question attached, a workbook nobody explained, one page for a manager who has twelve tabs of your analysis, a deck that has to land — and then the proof of it an employer can read in two minutes.
The skills follow the work in the order it happens, in seven stages, and each stage hands a named artifact to the next. Get AI-fluent for work hands over the delegation map, the Claude Project, the task brief, the output check and the use statement. Research the business hands over the company research brief, the PESTLE, the SWOT and the weekly commercial briefing. Analyse the numbers hands over the issue tree, the explained workbook, the analysis walkthrough, the financial model and the sanity check. Write for busy people hands over the executive summary, the email to a director, the business case and the weekly update. Present it hands over the storyline, the chart choices, the deck and the review. Build a proof project hands over the brief, the work log, the case study and the reflection. Show it to employers hands over the decoded job ad, the CV evidence lines, the interview answers, the voice check and the practice case.
The red line: Claude helps you prepare, practise, research, organise and critique your own work. It never sits a test, interview or assessment for you, never invents experience or results, and nothing goes to an employer that you have not checked and cannot explain yourself.
Each skill is one named method or one artifact, and keeps the real mechanics inside: the canvas blocks, the six PESTLE factors, branches that do not overlap, separate input and calculation tabs, assertion titles. It names the inputs it needs, where it breaks, a fixed output template, a finish line and the next skill to run. No invented numbers or experience — no made-up figures, quotes, results or projects; anything missing becomes a bracketed placeholder or an open question. Work is checked, people are not: the skills check evidence, options and fit to a brief, never the graduate or anyone else.
The first stage practises the four competencies of the AI Fluency framework — Delegation, Description, Discernment and Diligence — which the pack names and links rather than reproduces. Where a skill fits a Claude surface — Projects, web search, Claude in Chrome, Claude for Excel, Claude Docs, Claude Slides or Claude for PowerPoint — it says so, and beta features are marked beta.
Mechanically, each skill is one folder with a SKILL.md file. The 31 are grouped into the seven stages, and each one gives full value on its own: run one, run a stage, or work the whole thing through in the order it arrives.
Download all 31 skills. One zip: ready-to-install skill zips, readable SKILL.md files, and the sources 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 work, in seven stages: getting AI-fluent, researching the business, analysing the numbers, writing for busy people, presenting it, building a proof project — and showing it to employers.
1 · Get AI-fluent for work
1. AI Delegation Map
Use when: You use AI for everything or nothing and cannot say where it helps
Output: Your recurring tasks sorted into do it yourself, Claude helps, Claude drafts and you check, never AI, with reason, risk and the rules that apply
2. Claude Project Setup
Use when: Every chat starts from zero and you keep re-explaining your work
Output: Project instructions, files to add, a "never paste" list, a checking rule, three test prompts
3. AI Task Brief
Use when: Claude's first answer is generic because your request was one line
Output: Goal, audience, sources, format, an example of good, constraints, how Claude should work, the check you will run
4. AI Output Check
Use when: The answer sounds right and you are about to send it to someone senior
Output: Claim-by-claim table, numbers recomputed, reasoning gaps, fix or cut list, a revised brief
5. AI Use Statement
Use when: You are unsure whether or how to say you used AI
Output: What Claude did, what you did, what you checked, what data went in, in three lengths
2 · Research the business
6. Company Research Brief
Use when: You need to understand a company fast for a project, a call or an interview
Output: Business model canvas, revenue lines, customers, competitors, results and filings, recent news, every line sourced
7. PESTLE Analysis
Use when: A manager asks what is going on in the market and a list of trends will not do
Output: Six external factors for one sector, each with a dated source, likely effect and "so what", plus the one to watch
8. SWOT Analysis
Use when: You must present a SWOT and want more than four boxes of adjectives
Output: Evidence-backed strengths, weaknesses, opportunities and threats, cross-moves, two priorities
9. Commercial Awareness Briefing
Use when: You are asked what challenges the business faces and only have headlines
Output: Three stories a week, why each matters, "so what for this employer", a question to ask, sources
3 · Analyse the numbers
10. Issue Tree
Use when: You are handed a vague question and do not know where to start
Output: Problem statement, branches that do not overlap, a hypothesis per branch, the analysis each needs, the order
11. Spreadsheet Explainer
Use when: You inherited a workbook nobody explained and must update it by Friday
Output: Formula map, each step in plain words with its cell, a rebuild exercise, questions for the owner
12. Data Analysis Walkthrough
Use when: You have a raw export and a question and need findings you can defend
Output: Cleaning log, each step with its formula or pivot, findings you recompute, limits of the data
13. Simple Financial Model
Use when: Someone asks what it would cost or make and you need a model, not a guess
Output: Inputs, calculations and outputs on separate tabs, sourced assumptions, base and two scenarios
14. Spreadsheet Sanity Check
Use when: You are about to send numbers upward and one wrong cell would cost trust
Output: Cross-footed totals, units and signs, magnitude checks, broken references, hard-codes, pass, fix or ask list
4 · Write for busy people
15. Executive Summary
Use when: Your manager wants one page and your analysis is twelve tabs
Output: The answer first, three supporting points with evidence, the ask, one page, checked against the analysis
16. Email to Senior Stakeholders
Use when: You need something from a director and keep rewriting the first sentence
Output: A subject that states the point, bottom line first, two lines of context, the ask with a date, a tone check
17. Business Case
Use when: Your recommendation must survive the finance question
Output: Case for change, options in a trade-off table, sourced costs and benefits, pre-mortem risks, delivery, the decision asked
18. Weekly Status Update
Use when: You send a weekly update and nobody reads past the first line
Output: Done, next, blocked, RAG per workstream with the reason, asks of your manager
5 · Present it
19. Deck Storyline
Use when: You are about to open a blank deck and the story is not clear yet
Output: Situation, complication, resolution, an action title per slide, the evidence each needs, a slide budget
20. Chart Choice
Use when: Your slides have the right numbers and nobody can see the point
Output: Per number, the message, the chart or table that shows it, a takeaway title, axis rules, what to cut
21. Slide Deck Build
Use when: The storyline is agreed and you need the deck by tomorrow
Output: Slides from the storyline, assertion titles with evidence, speaker notes, a source line on every number
22. Deck Review
Use when: The deck is done and you are about to send or present it
Output: Titles-only read, a sceptic's read of numbers and sources, consistency check, an edit list by severity
6 · Build a proof project
23. Proof Project Brief
Use when: You have no work experience to point to and want something real to show
Output: One real business question from public data, the deliverable, a two-week scope, sources, what Claude will not do
24. AI Work Log
Use when: You want to show how you used AI, not just say it
Output: Dated entries of what you asked, what came back, what you checked and changed, and why
25. Portfolio Case Study
Use when: The project is finished and you need a version an employer reads in two minutes
Output: One page with the question, your approach, your decisions, what Claude did and what you checked, the result
26. Project Reflection
Use when: You finished something and want the lessons ready for interviews
Output: What happened, what worked, what you learned about the work and about AI, what you would change, written by you
7 · Show it to employers
27. Job Ad Decoder
Use when: The ad lists twelve skills and you cannot tell what the job is
Output: Must-haves, nice-to-haves, hidden tasks, AI and data skills, your evidence for each, gaps and what to do
28. CV Evidence Lines
Use when: Your CV says "proficient in AI" and so does everyone else's
Output: CV bullets with action, tool, how you checked it and result, each traceable to your work log
29. AI Interview Answer
Use when: An interviewer asks how you use AI and a list of tools will not do
Output: Two or three STAR stories on how you briefed, checked and disclosed AI, practised aloud with follow-ups
30. Application Voice Check
Use when: Your application reads like everyone else's AI-polished answer
Output: Generic phrases, claims you cannot back, missing examples, words you would never say, an edit list you apply
31. Assessment Centre Case Practice
Use when: You have an assessment centre next week and have never done a case exercise
Output: A fictional practice case, your notes and recommendation, a critique of structure and numbers, a presentation plan
Where to start, and how the skills chain
There is no compulsory sequence. Bring the thing that is due: if you have a company to understand by Thursday, start at Research the business; if you inherited a workbook, start at Analyse the numbers; if your manager asked for one page, start at Write for busy people; if the application is open, start at Show it to employers. Every skill lists the inputs it needs, so it tells you when the stage before it is the one that is actually missing.
They chain. gbiz-ai-delegation-map decides what Claude touches at all and gbiz-claude-project-setup stops every chat starting from zero; gbiz-ai-task-brief gets a first answer worth having, gbiz-ai-output-check reads it before anyone senior does, and gbiz-ai-use-statement says how you used it. gbiz-company-research, gbiz-pestle-analysis, gbiz-swot-analysis and gbiz-commercial-awareness build the picture of the business and its market. gbiz-issue-tree turns a vague question into branches, gbiz-spreadsheet-explainer and gbiz-data-analysis make a workbook and an export something you can defend, gbiz-financial-model builds the numbers and gbiz-sanity-check checks them before they go upward. gbiz-executive-summary, gbiz-email-to-senior, gbiz-business-case and gbiz-status-update put the answer in front of busy people; gbiz-storyline, gbiz-chart-choice, gbiz-slide-deck and gbiz-deck-review turn it into a deck with a story and a last read. gbiz-proof-project-brief gives you something real to show, gbiz-ai-work-log records how you worked, gbiz-portfolio-case-study and gbiz-project-reflection make it readable and reusable. Then gbiz-job-ad-decoder, gbiz-cv-evidence-lines, gbiz-ai-interview-answer, gbiz-application-voice-check and gbiz-case-study-practice take it to employers — each CV line and interview story traceable back to the work log.
The sources behind each method are in resources/evidence-and-sources.md, inside the pack.
Setup guide
- 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 resources/evidence-and-sources.md, which names the source behind every method and where it stops.
- Install your skills. In Claude Code, add the marketplace and install grad-business-pack, and all 31 load at once. In Claude, turn on code execution and file creation 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.
- Start with whatever is due this week. There is no setup phase to clear first. Pick the thing on your desk — a company to understand, a workbook to update, one page for your manager, a deck, an application, an assessment centre — and run the one skill named for it. Each skill lists the inputs it needs and names the next skill to run, so one run turns into the chain on its own.
Where to start
| Your situation | Skill to run |
|---|---|
| You cannot say what to give AI and what to keep? | AI Delegation Map |
| Every chat starts from zero? | Claude Project Setup |
| The answer sounds right and it is going to someone senior? | AI Output Check |
| A company to understand before Thursday's call? | Company Research Brief |
| A vague question and no idea where to start? | Issue Tree |
| A workbook you inherited and must update by Friday? | Spreadsheet Explainer |
| Numbers about to go upward, and one wrong cell would cost trust? | Spreadsheet Sanity Check |
| Your manager wants one page and your analysis is twelve tabs? | Executive Summary |
| A blank deck and no story yet? | Deck Storyline |
| No work experience to point to? | Proof Project Brief |
| A CV that says "proficient in AI", like everyone else's? | CV Evidence Lines |
| An assessment centre next week and no case practice? | Assessment Centre Case Practice |
The quality bar
Every skill in the pack holds the same standard, the one we hold when we run our own company:
- One method, applied properly: each skill uses the real mechanics of its source — the canvas blocks, the six PESTLE factors, branches that do not overlap, separate input and calculation tabs, the five cases, assertion titles, the six reflection stages — not a generic gather, analyse, recommend
- Every output gets checked: claims trace to a source, numbers are recomputed, and every group has a skill that reads the work before it goes to someone senior
- No invented numbers or experience: no made-up figures, quotes, results or projects — anything missing becomes a bracketed placeholder or an open question
- You do the thinking you will be asked about: Claude explains its working so you can rebuild the key steps and answer the follow-up questions yourself
- Work is checked, people are not: skills check evidence, options, work and fit to a brief, never the graduate or anyone else. Legal, data protection and advertising points end with "check with a qualified adviser"; course work stays within your university's rules and work tasks within your employer's AI policy
- The red line: Claude helps you prepare, practise, research, organise and critique your own work — it never sits a test, interview or assessment for you, never invents experience or results, and nothing goes to an employer that you have not checked and cannot explain yourself
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 work a business graduate is hired to do. When you want a whole team working this way, AI carrying the overhead so people do the part only people can do, across analysis, writing and everyday operations, that's what we build with clients.
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 grad-business-pack@polar-bear-skills
Using Claude on the web instead? Download the zip and upload each skill from its install folder under Customize → Skills.