gpai-pflichten
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE]: The skill consists entirely of markdown instructions and metadata for an AI agent to perform legal analysis. There are no scripts, binaries, or automated execution steps involved.
- [SAFE]: All external links point to 'eur-lex.europa.eu', which is the official website for European Union law and a trusted source for legislative text. The skill uses standard markdown and YAML frontmatter without any obfuscation or suspicious patterns.
- [METADATA_POISONING]: The metadata fields (name, description, tags) are descriptive and aligned with the functional purpose of the skill (legal compliance for GPAI models). No deceptive or malicious instructions were found in the metadata.
- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process external inputs (model types, training data descriptions), it includes internal warnings (e.g., regarding the lack of an official list of systemic models) and follows a structured output format, which helps maintain context and prevents accidental obedience to instructions that might be embedded in user-provided case facts.
Audit Metadata