dynamic-workflow-mode
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes workflows that process external and potentially untrusted data sources.
- Ingestion points: SKILL.md identifies inputs including external files, URLs, and prompts.
- Boundary markers: Absent; the provided harness templates do not explicitly instruct the agent to use delimiters or ignore instructions within these inputs.
- Capability inventory: SKILL.md defines capabilities for the agent to generate new artifacts and execute evaluation commands within the workspace.
- Sanitization: Present; the skill recommends safety checks such as 'config validation' and 'no-secret scans'.
- [DYNAMIC_EXECUTION]: The skill guides the agent in generating and executing task-specific automation scripts (harnesses).
- Evidence: The 'Loop' and 'Task-Local Harness Template' sections in SKILL.md outline a process for generating artifacts and running evaluation commands at runtime.
Audit Metadata