open-websearch-maintainer
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides structured guidance for repository maintenance, focusing on consistency in networking logic and requiring test validation for all changes. No malicious patterns or unauthorized behaviors were detected.\n- [INDIRECT_PROMPT_INJECTION]: The skill involves processing and modifying a codebase, which is a standard surface for indirect prompt injection via external code or documentation. The instructions mitigate this by providing strict scope discipline and core rules for the agent to follow.\n
- Ingestion points: The agent interacts with repository files including source code (src/), tests, and documentation (README.md).\n
- Boundary markers: The skill provides 'Scope discipline' and 'Core rules' to constrain agent actions during the maintenance process.\n
- Capability inventory: Modifying local project files and executing standard development tools (npm, npx).\n
- Sanitization: The skill relies on specific procedural rules rather than automated content filtering.\n- [COMMAND_EXECUTION]: The skill guides the user to run local development and testing commands as part of the validation workflow.\n
- Evidence: References to npx tsc --noEmit and npm run test in references/validation.md for project verification.
Audit Metadata