brida-reflex-incoming-message-risk
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a framework for analyzing external messages which constitutes an attack surface for indirect prompt injection.
- Ingestion points: The skill ingests data via the
messageExcerptandsubjectfields as defined in thecustom-reflex.jsonfixtures andSKILL.md. - Boundary markers: The instructions in
custom-reflex.jsoninclude specific boundary directives: "Judge only the message context supplied; do not infer facts or authority that are absent." - Capability inventory: The skill focus is recommendation classification; no high-privilege capabilities such as arbitrary command execution are defined.
- Sanitization: The
playbook.mdguidance promotes using redacted or synthetic excerpts and supply only bounded context. - [EXTERNAL_DOWNLOADS]: The skill instructions in
SKILL.mddirect the agent to fetch configuration and availability data from live hosted files (/reflex/llms.txt,/reflex/agent.md, and/reflex/agent.json) associated with the vendor's infrastructure.
Audit Metadata