external-resource-context
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied text for resource locations and access commands and commits this data directly to the repository's documentation, which could be used to inject malicious instructions that influence downstream agents or users.\n
- Ingestion points: User answers to domain-specific axis questions and the self-declaration phase defined in
references/hearing.md.\n - Boundary markers: The output is structured in markdown, but the skill lacks explicit delimiters or 'ignore' instructions for the captured resource fields to prevent instruction interpretation.\n
- Capability inventory: The skill performs file writes to
docs/project-context/external-resources.mdand other feature-specific design documents.\n - Sanitization: No evidence of input validation, escaping, or filtering of user-provided content before it is committed to files.
Audit Metadata