graph-rag
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a procedure for ingesting external data (graph entities and relationships) which constitutes an indirect prompt injection surface.
- Ingestion points: Defined in Step 3 ('Expand With Bounded Traversal') and Step 4 ('Build Prompt Context') of
SKILL.md. - Boundary markers: The skill encourages the use of structured Markdown headers like
## Anchorsand## Supporting Evidenceto delimit retrieved content from instructions. - Capability inventory: The skill contains no executable scripts, tool definitions, or file-writing capabilities.
- Sanitization: There is no explicit instruction for content sanitization, but the recommended bounded traversal and structured evidence building mitigate accidental instruction following.
Audit Metadata