Multi-Agent Systems Architect
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes an architecture for agents that process untrusted external data such as web pages and documents, which represents an indirect prompt injection attack surface.
- Ingestion points: System instructions explicitly mention processing 'web pages, documents, or user input' in SKILL.md.
- Boundary markers: The instructions mandate isolating content from instructions and using explicit schemas to delimit data.
- Capability inventory: The design framework includes capabilities for subprocess execution (code execution), file system writes, and network operations (web search, APIs) across various agent roles.
- Sanitization: The skill advocates for 'sanitizer agents' to extract structured data and strict schema validation to mitigate injection risks.
- [SAFE]: The skill promotes secure engineering practices, including least-privilege tool access, audit logging for tool calls, and the use of sandboxed environments for code execution.
Audit Metadata