omh-model-optimization
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute several CLI commands using the
omhtool suite (e.g.,omh coding model-route,omh model-chains set,omh runtime record) to manage model onboarding and optimization workflows. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill explicitly requires the agent to research and process external data from official release documentation and community harness findings (referenced in SKILL.md).
- Boundary markers: The instructions specify that research findings must be labeled by source (e.g., "official", "observed", "community") to help distinguish the provenance of information.
- Capability inventory: The agent has the capability to modify model routing configurations and prices via CLI tools (
omh model-chains,omh coding category-maestro) based on the ingested information. - Sanitization: While the skill emphasizes labeling and evidence-gated quality tiers, there is no technical sanitization described for the content being processed before it influences agent decisions.
Audit Metadata