brida-reflex-semantic-record-match
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interprets user-provided entity data (names, categories, and locations) to perform matching logic, which presents a surface for potential injection if malicious content is embedded within these fields to influence agent recommendations. \n
- Ingestion points: Runtime record data processed as defined in
SKILL.mdandreferences/playbook.md. \n - Boundary markers: Instructions mandate restricting results to specific declared branches and using only supplied comparison fields. \n
- Capability inventory: Integration is limited to recommendation-only via MCP, SDK, or REST calls, without permissions for mutations or protected actions. \n
- Sanitization: The playbook requires deterministic matching and exclusions to be processed before the semantic check, and mandates parsing in deterministic host code. \n- [EXTERNAL_DOWNLOADS]: Retrieves service availability, technical documentation, and configuration from the developer's official endpoints (e.g.,
/reflex/llms.txt,/reflex/agent.json).
Audit Metadata