maintain-loop
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes 'smoke tests' and 'Verify' checks as defined in the
loop.mdandrules.jsonconfiguration files. - [DYNAMIC_EXECUTION]: Executes commands or scripts based on dynamic rule definitions and metadata stored within the skill's data directory.
- [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: Processes data from external 'Source' locations and 'Gate' results which are potentially untrusted.
- Boundary markers: The instructions do not specify any delimiters or safety markers to isolate processed data from the prompt logic.
- Capability inventory: Includes shell command execution (subprocess) and the ability to apply results to external 'Sinks'.
- Sanitization: No explicit sanitization or filtering is described; security depends on the user's manual approval of generated proposals.
Audit Metadata