benchmark-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from third-party websites and user-supplied exports. It explicitly manages this risk through detailed instructions.
- Ingestion points: Web pages, CSV exports, and tool outputs are processed to extract metrics.
- Boundary markers: Instructions in references/agent-security.md define external content as data-only and require reporting injection attempts as findings.
- Capability inventory: The skill performs network fetches and file writes to the .agents/ directory.
- Sanitization: Emitted markup must be HTML-escaped, and sensitive attributes are restricted to prevent XSS.
Audit Metadata