external-resource-context
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the capture and persistence of user-supplied access methods (including shell commands, URLs, and file paths) into a centralized project-context file (docs/project-context/external-resources.md). These records are specifically intended to be used by subsequent agent sessions to retrieve external resource content, creating a path where malicious input could be executed or visited by future processes.
- Ingestion points: Interactive user responses provided during the multi-domain "hearing" process (defined in references/hearing.md).
- Boundary markers: The generated documentation lacks delimiters or instructional warnings to prevent the agent from obeying instructions potentially embedded in the recorded fields.
- Capability inventory: The skill performs file writes to create the docs/project-context/ directory and manage the external-resources.md file.
- Sanitization: There is no evidence of validation or sanitization for the captured access methods, such as checking for command injection sequences in recorded CLI commands.
Audit Metadata