brida-reflex-semantic-cache-reuse-gate
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references live configuration files (llms.txt, agent.md, agent.json) hosted on the vendor's (brida-ai) infrastructure to verify feature availability. These are standard vendor-provided metadata resources.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted request data entering the agent context via fixtures (references/custom-reflex.json) and runtime interaction. It mitigates this via an explicit authority boundary in SKILL.md that prohibits the result from granting permissions for sensitive actions. The capability inventory shows no dangerous subprocess or file-write commands across the scripts. The skill further enforces safety by restricting state ingestion to non-sensitive data classes.
Audit Metadata