brida-reflex-outbound-message-risk
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing untrusted data (message drafts) to evaluate semantic risk.
- Ingestion points: The
draftstate field used in the Reflex classification contract defined inreferences/custom-reflex.json. - Boundary markers: The
SKILL.mdandreferences/playbook.mdestablish an "Authority boundary" stating the Reflex is recommendation-only and cannot perform protected actions (payments, deploys, etc.). - Capability inventory: The skill facilitates semantic classification via the Brida Reflex SDK/API; it does not request capabilities to write files, execute shell commands, or perform mutations based on the untrusted input.
- Sanitization: The instructions mandate using a
non_sensitivedata class and explicitly prohibit placing credentials or personal data in the Reflex state. - [EXTERNAL_DOWNLOADS]: The skill refers to live configuration and documentation files hosted on the vendor's infrastructure to verify SDK and API availability.
Audit Metadata