ltx-director
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the AI agent to ingest, parse, and modify the
timeline_datawidget from external ComfyUI workflows, which creates a potential vector for malicious data to influence the agent's behavior.\n - Ingestion points: The agent retrieves untrusted workflow data containing hidden JSON strings through the
get_workflowtool, as documented in the procedure section ofSKILL.md.\n - Boundary markers: No explicit boundary markers or instructions to ignore potential commands within the
timeline_dataJSON are provided to the agent.\n - Capability inventory: The skill leverages tools like
set_inputandcreate_workflowto write modified states back to the application environment, allowing a potential payload in the input data to affect system configuration.\n - Sanitization: The instructions focus on structural validity (JSON parsing) but do not include content sanitization or validation of the segment prompts against injection patterns.
Audit Metadata