brida-reflex-memory-relevance-gate
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to consult live availability and SDK status files from the vendor's official resources (specifically
/reflex/llms.txt,/reflex/agent.md, and/reflex/agent.json). These are used for operational awareness and do not include instructions for remote code execution. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data (observations) to classify them for potential storage. It includes architectural safeguards to mitigate risks:
- Ingestion points: Data is sourced from the project environment or active event state as described in
SKILL.md. - Boundary markers: The
Authoritysection inSKILL.mdand theAuthority boundaryinreferences/playbook.mdexplicitly state that the logic is recommendation-only and never grants permission for sensitive or protected actions like payments or account changes. - Capability inventory: The skill performs verification via MCP, SDK, or REST calls using provided synthetic fixtures in
references/custom-reflex.json. - Sanitization: The workflow mandates deterministic privacy, secret, and tenant checks before data reaches the Reflex logic, ensuring that sensitive data is filtered out prior to processing.
Audit Metadata