agent-workflow-designer
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied JSON workflow specifications through utility scripts for estimation and validation. * Ingestion points: scripts/cost_estimator.py, scripts/multi_agent_cost_estimator.py, scripts/workflow_validator.py, and scripts/workflow_visualizer.py load data from local files or stdin. * Boundary markers: None identified, as the scripts are standard Python utilities rather than prompt templates. * Capability inventory: The scripts are limited to stdout reporting and data analysis; no network calls, file-write operations, or subprocess executions are performed. * Sanitization: Structural validation of the JSON input is performed by the scripts to ensure valid workflow structures.
- [SAFE]: No evidence of malicious behavior, obfuscation, or unauthorized access to system resources was found. The skill operates within a safe local context for design and estimation purposes.
Audit Metadata