benchmark
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown documentation without any associated scripts, executables, or configuration files that could perform malicious actions.
- [DATA_EXPOSURE]: The skill describes a workflow for saving performance baselines to a
.benchmarks/directory. This is standard behavior for performance tools and does not constitute a risk of unauthorized data exposure. - [INDIRECT_PROMPT_INJECTION]: The skill outlines tasks that involve processing external data (navigating to URLs and hitting API endpoints). While this creates a theoretical surface for indirect prompt injection from those external sources, the skill itself does not implement or encourage unsafe processing of that data.
- Ingestion points: Target URLs and API response bodies (SKILL.md).
- Boundary markers: None specified in the documentation.
- Capability inventory: Browser navigation, HTTP requests, and execution of local build tools (TypeScript, Docker, etc.).
- Sanitization: Not specified; assumes standard agent-level safeguards.
Audit Metadata