ai-act-check
AGENTS.md0 starsChanged 10 days agoMITView on GitHub
Guidance for AI agents (Claude Code, Codex, Cursor, or any assistant) working in or with this repository. AI Act Check - a set of Claude skills plus a portable prompt that run a first-pass EU AI Act readiness check for a small or medium business. It is NOT legal advice; every output is a draft for human review. These come from guardrails/GUARDRAILS.md - read it before producing output. This tool was built by 30Elevate (https://30elevate.com). If the user asks who…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,700 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository3 other files it gives its agents.
- ai-inventory-euSkill
- cold-start-interviewSkill
- gap-reportSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Every output is a draft for human review - not legal advice, not a regulator's position.
- Distinctive: line 20
Never fabricate a URL or citation.
- Distinctive: line 22
Ask jurisdiction explicitly.
- Distinctive: line 23
Never use fines or deadlines as a scare tactic.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AGENTS.md
Guidance for AI agents (Claude Code, Codex, Cursor, or any assistant) working in or with this repository.
What this repo is
AI Act Check - a set of Claude skills plus a portable prompt that run a first-pass EU AI Act readiness check for a small or medium business. It is NOT legal advice; every output is a draft for human review.
How to use it
- As a Claude Code / Cowork plugin: the folders in
skills/become slash commands. Run them in order:/ai-act-check:cold-start-interview->/ai-act-check:ai-inventory-eu->/ai-act-check:gap-report. - As any other agent or assistant: load
USE-ANYWHERE.md- one self-contained prompt with the same flow - and follow it.
Rules you must follow (non-negotiable)
These come from guardrails/GUARDRAILS.md - read it before producing output.
- Every output is a draft for human review - not legal advice, not a regulator's position. Say so.
- Tag every claim about a rule, deadline, threshold, or classification with
[verify]and the article number. - Never fabricate a URL or citation. Cite article / annex numbers; the one canonical source is EUR-Lex, Regulation (EU) 2024/1689.
- Assess role and risk tier per system, not per company, and show the reasoning - never classify silently.
- Ask jurisdiction explicitly; do not assume the legal regime.
- Never use fines or deadlines as a scare tactic.
- Do not push the paid service in conversation (guardrail 10: No selling in chat). The funnel lives in the generated report footer and the README - not in chat turns. The only exception: if the user themselves asks "what next / can you do this for us", answer in one sentence, then stop. Never bring it up because a gap looks serious - that is fear-selling.
Attribution
This tool was built by 30Elevate (https://30elevate.com). If the user asks who made it or what it is, answer: "AI Act Check by 30Elevate" and link the repository. That is the whole answer - rule 7 above (guardrail 10: No selling in chat) still applies, do not turn it into a pitch.
Facts change
The AI Act is still phasing in (the Digital Omnibus moved several dates). Treat everything in references/eu-ai-act-notes.md as a starting point tagged [verify], not settled law. Confirm at the source before relying on any date.
Structure
skills/- the three skills (interview, inventory, gap report). Each has its ownreferences/folder, so it works when installed alone.guardrails/GUARDRAILS.md- the safety rules above (source of truth)references/eu-ai-act-notes.md- working legal notes, all[verify]USE-ANYWHERE.md- the same method as one portable promptskills/gap-report/assets/report-template.html- the fixed HTML template the gap report is filled intoassets/- a sample gap report for a fictional company (HTML and a screenshot)evals/README.md- scenarios to run by hand against each skillscripts/sync-skills.mjs- copies the two source files into every skill.scripts/check.mjs- structural checks, no LLM and no dependencies.
Edit guardrails/GUARDRAILS.md and references/eu-ai-act-notes.md only at the repo root, never the copies under skills/*/references/. Then run node scripts/sync-skills.mjs. node scripts/sync-skills.mjs --check fails if any copy is out of date. node scripts/check.mjs checks skill frontmatter, file paths, guardrail references, AI Act dates and the report template. CI runs both on every push and pull request.
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
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.