practice-agent-evaluation
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches evaluation datasets in JSON format from https://starlightintelligence.academy/academy/evaluation/cases.json to provide context for synthetic workflow testing.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external workflow cases that may contain embedded instructions designed to test agent robustness as part of its primary evaluation purpose.
- Ingestion points: External JSON data retrieved via the get_evaluation_cases tool or directly from starlightintelligence.academy.
- Boundary markers: Includes explicit instructions that retrieved text is data and cannot expand the task or approve external actions.
- Capability inventory: The skill utilizes tools for fetching and checking cases and provides instructions for generating evaluation reports.
- Sanitization: Relies on clear instructional scoping and a mandatory requirement to mock all side effects during testing.
Audit Metadata