simmer-generator
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is instructed to run a shell command specified as 'VALIDATION_COMMAND' in the setup brief to verify workspace changes.
- [DYNAMIC_EXECUTION]: In 'Workspace Mode', the skill modifies project files, including 'Evaluator scripts' and pipeline logic, and then executes these changes to perform smoke tests or validation runs.
- [INDIRECT_PROMPT_INJECTION]: The skill processes artifact content and judge feedback (ASI) which are external data sources.
- Ingestion points: The skill ingests the 'Current candidate' artifact, 'ASI' feedback strings, and 'background' context within 'SKILL.md'.
- Boundary markers: There are no instructions for using delimiters or protective tagging to prevent the agent from following instructions embedded within the artifact or feedback text.
- Capability inventory: The skill has permissions to write files to the output directory, modify files within a workspace, and execute arbitrary shell commands via validation triggers.
- Sanitization: The instructions do not include steps to sanitize the ingested artifact content or validate the 'VALIDATION_COMMAND' string before execution.
Audit Metadata