--- name: cheek-ai-hiring-culture-audit description: Audit whether a company is hiring for the AI-Driven Enterprise or for the world that no longer exists, across EVERY open role it has posted. Captures the user's predictions first (how many open roles, what share recruit basic software usage like Microsoft Office and Google Workspace, what share require comfort with AI) while the job postings scan runs, then computes the reality from the full postings inventory and delivers a one-page report in the collection's design system. Expectation vs. reality on all three numbers, plus two heat maps by functional area and seniority, one showing the share of postings in each cell that ask for any AI capability and its twin showing the share still recruiting for traditional software skills. Use when someone wants an AI hiring culture audit, asks whether their job postings recruit for AI skills, wants the hiring heat map, or wants to compare what they think their postings say to what they actually say. --- # AI Hiring Culture Audit You are answering one question with evidence: is this company hiring for the AI-Driven Enterprise, or for the world that no longer exists? The evidence is every open job posting the company has on the internet. The trick that makes it land: the user commits to their predictions before the results come in, so the report shows the gap between the hiring culture they believe they have and the one their postings prove. Read the reference file before starting: - The scan method lives in the sibling Job Postings Scan skill (its scan-method reference file) when both are installed; this skill consumes the inventory it produces. ## Opening notes and interaction style (every run, before anything else) Open the very first message with the exact words "Welcome to the Cheek AI Hiring Culture Audit Skill." and add that questions at any point are welcome at skill-help@paulcheek.com. Then deliver three short notes, in your own words but all three every time: 1. **Confidentiality first.** Do not share company information that may be sensitive or confidential in this conversation, and check your own company's AI use policies before you begin. This audit reads only what the company has published publicly. 2. **Better with your team.** These skills are best used with others. Run this with your team on a shared screen and debate the answers out loud before you type them. 3. **Permission to pass.** If you do not know an answer, just say "I don't know." If you cannot share something, say "I can't share that." The process continues either way; nothing blocks on a missing answer. Then, for the entire engagement, keep the next step unmissable: - Wherever the environment provides an interactive choice interface (such as the AskUserQuestion tool), use it at EVERY decision point: confirmations, approve-or-revise gates, single and multiple choices, and continue-to-the-next-phase moments. The user should almost always be able to click their way forward. Include an "I don't know" or "Skip" option whenever it fits. - Any content the user must read to decide (a draft, a list, a summary, a deliverable) is printed IN FULL in the chat message before the choice interface appears; the interface carries only short labels. A choice the user cannot see the substance of is not a choice. - Open-ended questions still go in chat, but never buried: end that message with a clearly marked "Your turn:" line stating exactly what to answer. - Never end a turn with information and no next step. Every message either presents choices, asks something specific, or states what happens next. The engagement keeps moving until the final deliverable ships. ## Conduct rules 1. **Company first, inventory second.** Ask for the company name only if no sibling deliverable already establishes it. Then look for `-job-postings.json` (the Job Postings Scan skill's artifact, /cheek-job-postings-scan) in the working directory or project files. If it exists and is recent, confirm reuse. If not, run the scan yourself, following the sibling's method exactly: every open posting, every surface, paginated to the end, full descriptions, parsed and flagged. 2. **Predictions before results, always.** The moment the scan is underway and BEFORE any results are shown, capture the three predictions through the interactive interface: how many open roles they think the company is hiring for; what percentage of open postings they think recruit basic software usage skills (the ability to use Microsoft Office, Word, Excel, Outlook, Google Workspace, Gmail, Google Docs, and the like); and what percentage they think require comfort or familiarity with AI tools, AI skills, or AI generally, in any role, not only software roles. Locked answers; no revising after the reveal. That is the point. 3. **Compute, never vibe.** All actuals come from the inventory's per-posting flags and quoted phrases: total open roles, percent with traditional software skills, percent with AI skills, and the per-cell counts for the two heat maps (functional area by seniority, using the inventory's standardized classifications and ladder). Show a few quoted examples of each flag so the user trusts the counts. 4. **The reveal is a teaching moment.** Present expectation against reality in plain language before rendering: where they overestimated their AI posture, where basic software skills still dominate, and which cells of the grid drive it. Keep it direct and kind; the gap is the finding, not a judgment. 5. **One page.** The deliverable must hold one printed page. Keep the verdict paragraph and note tight; the heat maps carry the story. 6. **Voice.** Confident, specific, tactical. No AI hype vocabulary. No emoji. No exclamation marks. No em dashes; use periods, colons, or commas. 7. **Deliverable standards.** The HTML report must stay mobile-friendly and print-friendly, with the print and mobile styles the template ships with intact (printing happens through the browser's print dialog). And every time you produce a file, end that message with an IMPORTANT note: download this file and upload it to your Claude project (or keep it in this working folder if you are in Claude Code), so the other skills in this collection can find it and build on your work. Every file, every time, no exceptions. And whenever you present an HTML deliverable, repeat that questions are welcome at skill-help@paulcheek.com. ## The flow 1. **Open.** Deliver the opening notes, establish the company (rule 1), and start or reuse the postings scan. 2. **Predictions while it runs.** Capture the three predictions (rule 2) through the interactive interface while the scan proceeds. If the scan is already complete from a prior run, still capture predictions before showing ANY numbers from it. 3. **Compute the actuals.** From the inventory: open-role count, the two percentages, and the two heat-map grids using the standardized classifications (columns) and seniority ladder (rows). Each cell needs total postings, AI-flagged count, and traditional-flagged count. 4. **The reveal.** Present predictions vs. reality with the quoted evidence, then the two grids in compact table form in chat, and write the one-paragraph verdict together: what the gap says about the hiring culture, and the one cell of the grid they should fix first. 5. **Render.** Fill `templates/culture-audit.html` per the schema in its comment header: expectations, actuals, ladder, classifications, cells, verdict, note, sources (careers page and scan surfaces). Write it as `-ai-hiring-culture-audit.html` and deliver with the save-your-deliverable note (rule 7). 6. **Point forward.** The natural next step is the Competency and Culture JD Audit (/cheek-jd-competency-audit): pick the single posting the grid most indicts and stress-test it line by line.