804-regulations-eu-nis2
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting and analyzing external evidence inventories provided by the user. While this creates a potential surface for indirect prompt injection, the risk is mitigated by explicit instructions throughout the skill that mandate the use of sanitized, maintainer-prepared inventories and strictly prohibit the processing of raw code, logs, configuration files, or other operational text.\n
- Ingestion points: System evidence inventories provided by the maintainer in SKILL.md and report generation workflows.\n
- Boundary markers: Present; instructions repeatedly emphasize the 'SANITIZED EVIDENCE ONLY' rule and 'NO RAW OPERATIONAL CONTENT' boundary.\n
- Capability inventory: The skill is restricted to analyzing provided facts and generating structured reports based on established NIS2 patterns.\n
- Sanitization: The workflow requires that all evidence be sanitized outside the agent context before being presented to the skill.\n- [NO_CODE]: The skill package consists entirely of markdown documentation, references, and report templates. It does not include any executable scripts, binaries, or other active code components, significantly reducing the attack surface.\n- [SAFE]: The skill references official EU documentation from the eur-lex.europa.eu domain to provide regulatory context. This is a well-known and trusted source for legal information, and the references are handled neutrally.
Audit Metadata