agent-harness-optimizer
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a documentation and pattern guide for performance optimization. It does not include any executable scripts that download external content or perform unauthorized system modifications.
- [COMMAND_EXECUTION]: The skill provides examples of shell commands for state management, specifically reading and writing session data to a local project directory (.omc/). These operations are restricted to project-specific metadata and are essential for the primary purpose of session persistence.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill suggests logging tool inputs and session summaries to the local project folder. This data is handled within the local environment and does not involve network exfiltration or access to sensitive system credentials like SSH keys or environment secrets.
- [INDIRECT_PROMPT_INJECTION]: The memory persistence mechanism involves re-injecting data from previous sessions stored in local files. While this creates a surface for indirect prompt injection if prior sessions handled untrusted data, it is a standard design for stateful agents and is implemented using project-local storage with no immediate execution of the stored data.
Audit Metadata