EU AI Act compliance scanning, audit chains, and evidence export — inside Claude Code.
claude plugin marketplace add airblackbox/air-blackbox-plugin
claude plugin install air-blackbox@air-blackbox-plugin| Command | What it does |
|---|---|
/air-discover |
Find AI components in your project and classify their EU AI Act risk level |
/air-comply |
Run a full compliance scan — 51+ checks across Articles 9-15 |
/air-replay |
Replay recorded traces to detect behavioral drift between model versions |
/air-evidence |
Export signed evidence packages as JSON or PDF for regulatory handoff |
These activate automatically when Claude detects relevant questions:
- compliance-scan — Triggers when you ask about compliance, scanning, auditing, or EU AI Act. Runs the scanner and walks you through results.
- interpret-results — Triggers when you have scan output and need help prioritizing fixes or understanding what each finding means.
An EU AI Act technical advisor that maps scan findings to specific articles, ranks remediation by enforcement risk, and helps you implement code fixes — with concrete examples, not abstract guidance.
Just ask a question:
"Is my project compliant with the EU AI Act?"
Claude automatically runs the scanner, interprets the results, and offers to fix the failures — all without you memorizing CLI commands.
Or use a slash command directly:
/air-comply
- Python 3.9+
pip install air-blackbox(installed automatically by the plugin if missing)- For replay: Go gateway binary and MinIO vault
- For evidence export: Go
evidencectlbinary
Apache-2.0