Agent Orchestration Patterns
PassAudited by ClawScan on May 1, 2026.
Overview
This is an instruction-only skill about multi-agent workflow design, with no code or install steps, but users should be aware that its examples route task content through multiple models/providers.
This skill appears safe as an instruction-only orchestration guide. Before using it with real work, confirm that any task data, drafts, and review outputs can be shared with every model or provider in the workflow, and keep high-sensitivity data out of cross-model examples unless your organization permits it.
Findings (1)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If the task or draft contains confidential information, the workflow could expose that information to additional model providers or agents.
The example sends task content, drafts, and review output between different models/providers as part of the orchestration pattern.
cross_check = await call_model( model="gpt-4o", prompt=f"""Independent review... TASK: {task} DRAFT: {draft} SELF-REVIEW: {self_review}Use these orchestration patterns only with data that is approved for the selected providers, and define clear boundaries for what each agent or model may receive.