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 Authority section in SKILL.md and the Authority boundary in references/playbook.md explicitly 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
Risk Level
SAFE
Analyzed
Sep 21, 2026, 08:21 PM
Security Audit — agent-trust-hub — brida-reflex-memory-relevance-gate