recherche-normes

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill retrieves data from the LibreJustice MCP server (librejustice.fr) and provides references to various official legal databases. These include French government sites (Légifrance, Ministry of Foreign Affairs), international bodies (United Nations, European Union), and specialized legal portals (HUDOC, Refworld). These references are appropriate for the skill's purpose and target reputable domains.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes external legal texts and administrative guidance.
  • Ingestion points: Untrusted data enters the agent context via the get_norm and search_norms tools, which fetch content from external databases.
  • Boundary markers: The skill instructions emphasize treating the article text (body) as the authoritative norm and distinguishing it from auxiliary commentaires or doctrine.
  • Capability inventory: The agent's capabilities are limited to searching and retrieving text; it lacks functions for file system modification or outbound network requests beyond the established toolset.
  • Sanitization: The agent is strictly instructed to quote provisions verbatim and verify validity dates, which serves as a functional check against malformed or instructional content within the legal data.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 02:12 PM