omni-inference
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references configuration and documentation from the author's official GitHub repository.\n
- Evidence: Links to the OmniRoute project on GitHub in SKILL.md.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines endpoints for searching and fetching external web content, creating an attack surface for indirect prompt injection.\n
- Ingestion points: The web/search and web/fetch endpoints in SKILL.md and endpoints.md ingest content from the public internet into the agent's context.\n
- Boundary markers: The documentation does not define specific delimiters or instructions to ignore embedded commands for external content.\n
- Capability inventory: The skill acts as an API gateway for inference services and does not provide direct file system or shell access tools.\n
- Sanitization: No specific input sanitization procedures for fetched web content are mentioned.
Audit Metadata