balise-web-legal
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to audit external websites and services, creating a surface for indirect prompt injection from untrusted data sources.
- Ingestion points: As detailed in references/audit-and-evidence.md, the skill instructs the agent to inspect web pages, forms, cookie storage, network requests, and third-party integrations (SDKs, pixels, scripts) from target products.
- Boundary markers: The instructions include procedural delimiters, requiring the agent to use [À CONFIRMER] for unverified findings and to follow a mandatory 7-question lock-in phase in SKILL.md to establish trusted context before analysis.
- Capability inventory: To perform its duties, the agent is expected to use browsing and data inspection capabilities, which involve processing content that could contain adversarial instructions.
- Sanitization: The skill relies on manual cross-referencing and source verification rather than automated sanitization of external content.
Audit Metadata