elysia
Pass
Audited by Gen Agent Trust Hub on May 8, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill demonstrates server-side patterns that interpolate untrusted data from request bodies and headers (e.g.,
body.provider) into response strings. This creates an indirect prompt injection surface where external input enters the agent context. Ingestion points: Data extracted from request objects inSKILL.md. Boundary markers: Examples lack delimiters to isolate untrusted content. Capability inventory: The skill facilitates building HTTP API handlers. Sanitization: No evidence of input validation or escaping for the interpolated variables. - [EXTERNAL_DOWNLOADS]: Includes reference links to the official GitHub repositories for Hono and Cloudflare, which are trusted development resources.
Audit Metadata