hypothesys
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to proactively verify premises by gathering facts from the environment. This creates an attack surface where a user-provided hypothesis can influence the agent to search for data in sensitive locations.
- Ingestion points: User-provided hypotheses and premises described in SKILL.md.
- Boundary markers: Absent. The instructions do not define limits on which parts of the filesystem or which tools are appropriate for 'fact finding'.
- Capability inventory: The skill explicitly references filesystem access, tool usage, and the ability to 'dispatch a sub-agent' to perform background tasks.
- Sanitization: Absent. There is no instruction to validate or filter the 'facts' found before they are incorporated into the agent's context or shared in the interview.
- [COMMAND_EXECUTION]: The skill instructs the agent to autonomously execute tools and explore the filesystem ('dispatch a sub-agent to find it; don't ask the user for anything you could look up yourself'). This directive encourages the agent to bypass user oversight for environment interactions during the 'interview' phase.
Audit Metadata