coval-compare-runs
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructs the agent to perform read-only comparisons between different versions of conversation simulations. It relies on standard CLI tools and API pagination for data retrieval without executing untrusted code.
- [INDIRECT_PROMPT_INJECTION]: The skill processes conversation logs and metric outputs, which are external data sources. While these ingestion points exist, the instructions explicitly command the agent to remain in a read-only state, focusing on computation and inspection rather than execution, which mitigates the risk of acting on malicious content embedded in the logs.
- [COMMAND_EXECUTION]: The skill utilizes a specialized CLI (e.g.,
runs get,metric-detail) to fetch data. These commands are descriptive of the skill's analytical purpose within the Coval environment and do not include arbitrary or dangerous system commands.
Audit Metadata