omni-model-builder

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it ingests external data from the Omni API through commands like omni models yaml-get. This data is then processed by the agent without explicit boundary markers to delimit untrusted content. Although the skill includes robust validation and a branching workflow, the lack of sanitization delimiters for ingested YAML content presents a standard low-severity risk factor.
  • [SAFE]: The skill mandates a secure development lifecycle by requiring all model modifications to be performed on isolated branches (omni models create-branch) rather than directly on the shared production model.
  • [SAFE]: A mandatory safety protocol is established in the 'Hard Stop' section, which strictly prevents the agent from merging or promoting changes without explicit, active consent from the user in the current session.
  • [SAFE]: The skill correctly scopes all external dependencies and resource links to official vendor-controlled environments, specifically the exploreomni.co domain and the official exploreomni/cli GitHub repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:45 PM
Security Audit — agent-trust-hub — omni-model-builder