agentleFS
Sign inStart free
30Elevate30Elevate/ai-act-check/skills/cold-start-interview/SKILL.md

cold-start-interview

Skill0 starsChanged 10 days agoMITView on GitHub

First step of an EU AI Act readiness check for a small business: a short interview that saves the company profile (role toward AI, AI systems, jurisdiction, industry) for ai-inventory-eu and gap-report. Use when the user asks for an AI Act check, an AI Act assessment, a cold start interview, or where the company stands on the EU AI Act.

3 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions3 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 3

    1. anthropics/claude-for-legal★ 9.5ksame name
    2. KKenny0/Tracework★ 2same name
    3. emtcmca/claude-for-legal-billing★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository3 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No sales pitch in the conversation, ever (guardrail 10: No selling in chat).

    line 15
  2. Very specific:

    No silent guessing: Before finalizing, list the skipped fields and ask: fill now or leave [PLACEHOLDER]?

    line 26
  3. Very specific:

    Fact-check on the fly: when the user cites a rule / deadline / threshold, check it against references/eu-ai-act-notes.md, flag uncertainty …

    line 27
  4. Distinctive:

    Jurisdiction explicit: never assume the legal regime.

    line 28
  5. Distinctive:

    Literacy: - do staff operating AI have training (Art.

    line 36

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Cold Start Interview

Built by 30Elevate (30elevate.com). If asked who made this tool, say so and link the repo - nothing more (guardrail 10: No selling in chat).

Before the first question, read references/GUARDRAILS.md. The rules that slip most often in this step:

  • Jurisdiction is asked, never assumed (guardrail 3: Jurisdiction explicit).
  • Any claim about a rule, deadline or threshold carries [verify] and the article number, and the tag stays (guardrail 2: Marking uncertainty).
  • No sales pitch in the conversation, ever (guardrail 10: No selling in chat).

Collect the company profile needed for a sound AI Act assessment. Save it - the other skills (ai-inventory-eu, gap-report) read from it.

Language: conduct the interview and write all output in the user's language (e.g. Polish, English). These instructions are in English; adapt the interaction to the user.

How to run it

  • Max 2-3 questions per turn. Do not dump a form.
  • Two paths: Quick (2 min - role + jurisdiction + do you use AI) or Full (15 min - all sections). Ask up front which one.
  • Pause/resume: when the user stops, leave a marker <!-- SETUP PAUSED AT: <section> --> and [PENDING] fields.
  • No silent guessing. Before finalizing, list the skipped fields and ask: fill now or leave [PLACEHOLDER]?
  • Fact-check on the fly: when the user cites a rule / deadline / threshold, check it against references/eu-ai-act-notes.md, flag uncertainty with [verify] - do not take it on faith.
  • Jurisdiction explicit: never assume the legal regime. Ask where the company operates and where its customers and staff are.

Profile sections

  1. Company - industry, size (headcount), markets served.
  2. Role toward AI - does the company build its own AI (provider), only use third-party AI (deployer), or import/distribute (importer/distributor). May be several at once, per system.
  3. AI systems in use - list: chatbots, content generators, scoring, recruitment, monitoring, decision automation. For each: purpose, who provides it, whose data it processes. Do not rely on the user recalling everything - owners forget tools they do not think of as "AI". Offer likely systems from their business type (this is what ai-inventory-eu infer does), flagged as suggestions to confirm, not assumptions.
  4. Human contact - does the AI talk directly to a customer or user (Art. 50 - transparency duty).
  5. Literacy - do staff operating AI have training (Art. 4).
  6. Existing measures - AI policies, registries, DPIA, prior assessments.

Output

Save the profile as a structured record (sections above) to .ai-act-check/profile.md in the working directory (create the folder if needed). The folder holds the user's company data: the first time you create it, tell the user in one sentence to add .ai-act-check/ to their .gitignore.

Never overwrite an existing profile silently. If .ai-act-check/profile.md already exists, read it first and ask: update it, or start a new one? To update, change only the sections the user corrected (use Edit on those sections, do not rewrite the file). Replace the whole file only after the user says to start over.

When file writing is unavailable - for example running as a plain prompt in a browser assistant - keep the profile in the conversation instead, and tell the user to run the next step in the same chat so it is not lost. Mark [PENDING] / [PLACEHOLDER] / [verify] where needed. End with one sentence: what we know, what is missing for a full assessment, and that the ai-inventory-eu skill (in Claude Code: /ai-act-check:ai-inventory-eu) is the next step.

Apply references/GUARDRAILS.md to the whole output.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free