context-engineering
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is educational in nature, providing best practices for structuring information to improve AI agent performance and accuracy.
- [INDIRECT_PROMPT_INJECTION]: The instructions identify the risk surface for indirect prompt injection and provide mitigation strategies.
- Ingestion points: Identifies user-submitted content, third-party API responses, and external documentation as untrusted data sources.
- Boundary markers: Recommends that instruction-like content within these data sources should be treated as data to be surfaced to the user, rather than directives to be followed.
- Capability inventory: Mentions the use of Model Context Protocol (MCP) servers for filesystem, database, and GitHub access, which are components of the capability surface.
- Sanitization: Advises on the necessity of manual verification before acting on content from untrusted sources.
Audit Metadata