omh-internal-audit
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill establishes a manual, human-in-the-loop audit process. It contains no instructions for the agent to autonomously access ledgers, databases, or sensitive file systems. All audit data is expected to be provided by the operator or observed from verified output.
- [INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface as it is designed to ingest and analyze untrusted external data (audit evidence such as logs, documents, and approvals).
- Ingestion points: Evidence provided by the user or control owner for each sample item, as described in
SKILL.mdunder 'Required inputs'. - Boundary markers: The instructions explicitly require the agent to separate 'evidence obtained' from 'evidence described' and to verify all results against observed output rather than assertions.
- Capability inventory: The skill's operations are limited to reading reference files (
skill_view) and status logging (omh runtime). It lacks capabilities for arbitrary shell execution, network exfiltration, or writing to the host file system. - Sanitization: No specific sanitization or filtering of input evidence is mentioned; however, the skill's purpose is to evaluate this data rather than execute it.
Audit Metadata