brida-reflex-model-effort-routing
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted task descriptions and work characteristics to provide model tier recommendations.
- Ingestion points: Reads project state and task context from the environment as described in SKILL.md.
- Boundary markers: Explicitly restricts data state to non_sensitive classes and defines a clear authority boundary in references/playbook.md that prevents recommendations from authorizing sensitive actions.
- Capability inventory: Limited to semantic classification and recommendation; possesses no capabilities for destructive mutations, payments, or merged code as stated in SKILL.md and references/playbook.md.
- Sanitization: Relies on state normalization as described in the autonomous execution steps of SKILL.md.
- [EXTERNAL_DOWNLOADS]: The skill references live availability data from vendor-managed resources (e.g., /reflex/llms.txt, /reflex/agent.json) from the Brida live environment.
Audit Metadata