six-thinking-hats
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs subagents to use retrieval tools to ingest untrusted external data and process it within the session context.
- Ingestion points: User-provided TOPIC and CONTEXT in SKILL.md, as well as external content and evidence fetched by subagents using retrieval tools during Stage 2.
- Boundary markers: The instructions for subagents in Stage 2 use bolded variable placeholders for focus questions and hats, but they do not provide explicit instructions or delimiters to ignore or escape potentially malicious commands embedded in the retrieved data.
- Capability inventory: The skill has the capability to read external data via retrieval tools and write generated markdown reports to the local file system.
- Sanitization: There is no mechanism described for sanitizing, validating, or filtering external content before it is interpolated into the prompts for the subagents.
Audit Metadata