omh-v18-loop-taxonomy
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a documentation-oriented reference that provides architectural context for the agent. It describes structural mappings between OMH skills and Saturate loop kinds.
- [DATA_EXPOSURE]: The instructions reference local file paths for architecture documentation (e.g.,
~/src/witt3rd/saturate/ARCHITECTURE.md) and state persistence (.omh/state/). These references are consistent with the developer-focused use case and the provided author context ('witt3rd'), representing standard local project structure rather than unauthorized sensitive data access. - [INDIRECT_PROMPT_INJECTION]: The skill defines how the agent should interpret its iteration semantics and terminal states. While these instructions influence agent behavior, they are provided as a formal taxonomy and do not incorporate untrusted external data into the prompt context.
Audit Metadata