create-workflow-command
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill sets up an orchestration pattern where untrusted input from user arguments is forwarded to downstream sub-agents, establishing a potential injection path.
- Ingestion points: User arguments are captured via
$ARGUMENTSin the generated orchestrator command file (SKILL.md). - Boundary markers: The blueprint utilizes markdown code block delimiters (e.g.,
text $ARGUMENTS) to separate data from instructions. - Capability inventory: Sub-agents executing the task workflow maintain access to powerful tools such as Read, Write, and Grep/Glob.
- Sanitization: No active content filtering, prompt escaping, or strict validation schema is enforced on the interpolated arguments before they are evaluated by sub-agents.
Audit Metadata