legal-tos-privacy
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill instructs the agent to audit the codebase for sensitive data collection points, which involves scanning for authorization headers and tokens.
- Evidence: SKILL.md Phase 1.1 explicitly directs the agent to search for strings like 'Authorization', 'bearer', 'token', 'cookie', and 'session' to document data collection points.
- Evidence: While intended for identifying data flows to build a privacy policy, this logic requires the agent to interact with security-sensitive application logic and project configuration metadata.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the user's project files and marketing materials to generate legal documents, creating an attack surface for malicious instructions hidden in those files.
- Ingestion points: Project codebase files (Phase 1) and marketing/website materials (Phase 2).
- Boundary markers: None. The instructions do not include delimiters or directives to ignore instructions embedded within the files being audited.
- Capability inventory: File reading and searching capabilities provided by the agent's built-in tools.
- Sanitization: None. External content is processed and interpolated directly into the drafting process without validation or filtering.
Audit Metadata