菜单
Plan a complex task, delegate steps to multiple workers, synthesize results. Use when the task requires multiple specialists working sequentially or with dependencies. Syner's most powerful multi-agent pattern.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Entry point for any task. Detects user intent, classifies complexity, and decides the next step (respond, execute, plan, or delegate). Use as first step before invoking specific workflows.
Formal audit skill based on ISO 27001 and ISACA/ITAF standards. Use when a formal, structured audit report is needed — not for informal code reviews. Produces structured output with frontmatter and standard sections.
Generate output, evaluate against quality criteria, optimize iteratively until threshold is met. Use when quality is critical and first attempts may not be good enough.
Split a task into independent subtasks, execute all concurrently, merge results. Use when subtasks don't depend on each other and speed matters.
Classify input and delegate to a single specialist. Use when the task needs one specific handler — not multiple. Syner's simplest orchestration pattern.
This skill should be used when coordinating work across multiple specialists or domains. Triggers on multi-step tasks, cross-domain work, planning complex implementations, or resolving cross-cutting concerns.
| name | orchestrate |
| description | Plan a complex task, delegate steps to multiple workers, synthesize results. Use when the task requires multiple specialists working sequentially or with dependencies. Syner's most powerful multi-agent pattern. |
Break down complex tasks, delegate to specialists, synthesize results.
input → plan() → steps[] → delegate(step)[] → results[] → synthesize() → output
Plan: Break the task into steps
dependsOn other stepsgoalDelegate: Execute each step with the right worker
WorkerResult for each step (success/failure + data)Synthesize: Combine all results into final output
| Worker | Agent/Tool | Capabilities |
|---|---|---|
ecosystem | ecosystem agent | Extensions, integrations, vendor work |
osprotocol | osprotocol agent | Protocol schema, validation, domains |
code | Claude (direct) | Implementation, tests, refactoring |
research | Explore agent | Codebase search, architecture analysis |