moai-harness-learner
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the 'moai' CLI tool, which is a vendor-owned utility associated with the author 'modu-ai', to manage system state, fetch update proposals, and execute modifications. These commands are scoped to the project environment.
- [DATA_EXPOSURE]: The skill reads and writes state information, update proposals, and configuration files within the '.moai/' and '.claude/skills/' directories. This access is necessary for the orchestration of the learning subsystem's lifecycle.
- [INDIRECT_PROMPT_INJECTION]: The skill processes JSON data generated by the CLI tool containing proposed modifications for other skills. Potential risks from malicious data are mitigated by the '5-Layer Safety Pipeline', specifically the L5 Human Oversight requirement which surfaces all changes for manual user approval via 'AskUserQuestion'.
- [DYNAMIC_EXECUTION]: The skill facilitates the modification of other skill definitions (auto-updates). This capability is restricted by a 'Frozen Guard' safety layer that prevents modification of core agent paths and requires explicit user consent for changes to user-area skills.
Audit Metadata