c-level-advisor
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill suite is designed to ingest and process untrusted project-specific data to provide tailored executive advice.
- Ingestion points: The
context-engineandcs-onboardskills create and read from~/.claude/company-context.md. Thedecision-loggerandboard-meetingskills ingest data frommemory/board-meetings/decisions.mdand various project files like ADRs and OKR JSON files. - Boundary markers: While the instructions define structured formats for inter-agent communication and data files, there are no explicit instructions for the LLM to ignore or sanitize embedded natural language instructions within the data it processes (e.g., within business descriptions or meeting notes).
- Capability inventory: The skill allows for the execution of 25 Python scripts which perform complex calculations and data transformations. It can also invoke other agent skills via the router.
- Sanitization: No evidence of active sanitization or filtering was found for content read from the
company-context.mdor other data inputs before they are interpolated into the agent's reasoning context.
Audit Metadata