EU AI Act Assessment
Assess EU AI Act compliance for "$ARGUMENTS". Classify AI systems by risk level, evaluate prohibited practice avoidance, assess high-risk system requirements, and review transparency obligations.
Arguments. $ARGUMENTS may contain the project name and optional flags. Take the first token as <project> and treat --skip-checks as a flag; never include flags in file paths. All paths below use <project>. If --skip-checks is present, skip the prerequisite check and, before writing any artifact, append one line to .singularskills/log.jsonl: {"ts":"<ISO timestamp>","skill":"compliance-eu-ai-act","workflow":"compliance","phase":1,"project":"<project>","event":"skipped","artifact":"01-assess/","note":"<reason or \"none given\">"}.
Prerequisites
Part of a workflow. This skill is phase 1 (Assess) of the GRC Lifecycle workflow (compliance); cross-domain workflows run it too. First check .singularskills/log.jsonl for an open run for this project: a run-started line with "project":"<project>" and no later run-finished line for the same workflow and project. If one exists, the conductor is running you: continue here. Otherwise, if the user asked to start, run, continue or resume a process or workflow rather than this one step, do not continue here: invoke the workflow skill with the workflow the user asked for, or compliance when they named none, followed by <project>; it runs this phase in order, asks before it writes and records the run. Otherwise continue here: a later phase runs on its own, behind its gate.
Run this skill's gate before writing: sh <this skill's folder>/scripts/gate.sh <project> <project root>, where <project root> is the folder that contains .singularskills/ (use . when you are already there). If it exits non-zero, stop and tell the user what it printed, unless --skip-checks was given (see Arguments). On a refusal, before you stop, append this line to .singularskills/log.jsonl (create the folder and file if missing): {"ts":"<ISO timestamp>","skill":"compliance-eu-ai-act","workflow":"compliance","phase":1,"project":"<project>","event":"refused","artifact":"01-assess/","note":"<what the gate printed>"}. Once the gate has passed, or under --skip-checks once you have appended the skipped line, append this line before you start work: {"ts":"<ISO timestamp>","skill":"compliance-eu-ai-act","workflow":"compliance","phase":1,"project":"<project>","event":"started","artifact":"01-assess/","note":""}.
Read .singularskills/compliance/<project>/00-scope.md. If this file does not exist, tell the user:
Phase 0 (Scope) has not been completed for "". Run
/compliance-regulatory-landscape <project>first, or use--skip-checksto bypass.
Process
Read context files:
- Read
references/grc-lifecycle-guide.mdfor GRC methodology reference - Read
references/assessment-template.mdfor output structure - Read
.singularskills/compliance/<project>/00-scope.mdfor scope and control framework context
- Read
AI system inventory and risk classification:
- Inventory all AI systems deployed or under development
- Classify each system by EU AI Act risk level:
- Unacceptable risk (prohibited): social scoring, real-time remote biometric identification (with exceptions), manipulation/exploitation, emotion recognition in workplace/education
- High-risk: Annex III systems (biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, justice) and safety components of Annex I products
- Limited risk: transparency obligations (chatbots, emotion recognition, deepfakes, AI-generated content)
- Minimal risk: voluntary codes of conduct
- Document classification rationale for each system
Prohibited practices assessment:
- Check all AI systems against the prohibited practices list (Art 5)
- Assess subliminal manipulation techniques, exploitation of vulnerabilities
- Review social scoring systems or behavioral prediction for detrimental treatment
- Evaluate real-time remote biometric identification usage and exemptions
- Document any systems requiring immediate remediation or discontinuation
High-risk system requirements (if applicable):
- Risk management system (Art 9): Assess continuous risk identification, estimation, evaluation, and mitigation throughout the AI lifecycle
- Data governance (Art 10): Evaluate training, validation, and testing data quality, relevance, representativeness, and bias examination
- Technical documentation (Art 11): Review documentation completeness for demonstrating conformity
- Record-keeping (Art 12): Assess automatic logging capabilities for traceability
- Transparency to users (Art 13): Evaluate instructions for use, intended purpose, accuracy levels, and known limitations
- Human oversight (Art 14): Assess human-in-the-loop, human-on-the-loop, or human-in-command measures
- Accuracy, robustness, cybersecurity (Art 15): Evaluate performance metrics, resilience to errors, and security measures
General-purpose AI model obligations:
- Determine if any GPAI models are used or provided (Art 51-56)
- Assess technical documentation and information sharing obligations
- Evaluate compliance with Copyright Directive obligations
- For GPAI with systemic risk: assess model evaluation, adversarial testing, incident reporting, and cybersecurity requirements
Transparency obligations:
- Assess AI interaction disclosure (users must know they are interacting with AI)
- Evaluate emotion recognition and biometric categorization disclosure
- Review deepfake and AI-generated content labeling
- Assess AI-generated content marking requirements
Conformity assessment and registration:
- Determine applicable conformity assessment procedure (self-assessment or third-party)
- Assess EU database registration requirements
- Evaluate CE marking requirements for high-risk systems
- Review EU representative appointment for non-EU providers
Timeline and phased enforcement:
- Map current compliance against enforcement timeline (prohibited practices: Feb 2025, GPAI: Aug 2025, high-risk Annex III: Aug 2026, high-risk Annex I: Aug 2027)
- Identify time-critical compliance activities
- Assess readiness for applicable enforcement phases
Write the artifact to
.singularskills/compliance/<project>/01-assess/eu-ai-act.mdfollowing the assessment template structure with:- AI System Inventory — complete inventory with risk classification per system
- Prohibited Practices — screening results and any systems requiring action
- High-Risk Requirements — assessment per Art 9-15 requirement for each high-risk system
- GPAI Obligations — general-purpose AI model compliance status
- Transparency — disclosure and labeling obligations coverage
- Conformity and Registration — assessment procedure readiness
- Enforcement Timeline — phased compliance readiness map
- Evidence Inventory — existing evidence and evidence gaps
- Remediation Priorities — ranked list of gaps to address
Output
The EU AI Act assessment written to .singularskills/compliance/<project>/01-assess/eu-ai-act.md. Present a summary to the user highlighting:
- Number of AI systems inventoried and risk classification distribution
- Any prohibited practice concerns requiring immediate action
- High-risk system compliance readiness score — score = (Compliant + 0.5 × Partial) ÷ applicable requirements (N/A excluded), as a percentage
- Top 3 gaps requiring remediation
Record the run. After you have written the artifact, append this line to .singularskills/log.jsonl (the started line of Prerequisites, with "event":"finished"): {"ts":"<ISO timestamp>","skill":"compliance-eu-ai-act","workflow":"compliance","phase":1,"project":"<project>","event":"finished","artifact":"01-assess/","note":""}.
Open the artifact with this frontmatter (agent is the agent you run in, or unknown; gate is passed when the gate ran clean and skipped under --skip-checks):
---
skill: compliance-eu-ai-act
workflow: compliance
phase: 1
project: <project>
started: <ISO timestamp>
finished: <ISO timestamp>
agent: <the agent you run in, or "unknown">
gate: <passed|skipped>
---
End the artifact with ## Decisions and ## Open questions, in this shape:
## Decisions
- <what was chosen> — <why> — <what was rejected>
## Open questions
- <what the next phase or a reviewer must settle>
End the artifact with this paragraph, verbatim:
Not legal or audit advice. This artifact supports preparation for compliance work; certification, attestation and regulatory conclusions require a qualified auditor or counsel.