JSON format

JSON supports the same schema as YAML. Use it when a workflow is generated by another tool or when your team prefers strict JSON files.

Example

{
  "harness": "claude-cli",
  "model": "sonnet",
  "steps": [
    {
      "name": "inspect",
      "prompt": "Inspect @src/server/ and explain the main request flow."
    },
    {
      "name": "review",
      "resumeSession": true,
      "prompt": "List the two most important weaknesses in that flow."
    }
  ],
  "finalStep": {
    "prompt": "Summarize the recommended next action."
  }
}

When to choose JSON

• another tool creates workflows for you
• you want a format with stricter editing rules
• you plan to validate or transform workflows in external tooling

Use the same workflow schema reference regardless of file format.

Parallel steps

JSON uses the same parallel queue-item shape as YAML. The members are flattened into the workflow's authoritative step numbering:

{
  "steps": [
    {
      "parallel": [
        {
          "name": "research-api",
          "prompt": "Research the API implementation."
        },
        {
          "name": "research-tests",
          "prompt": "Research the existing test strategy."
        }
      ],
      "failFast": true
    },
    {
      "name": "synthesize",
      "priorContext": [1, 2],
      "prompt": "Synthesize the independent research."
    }
  ]
}

Group members share one workspace and, when configured, one managed or named worktree for the run. They are intended for independent work; v1 has no per-member isolation or automatic merge. queue.maxParallelSteps bounds the group width, while 1 preserves sequential execution. Group members may enable the built-in ctrlyoke dynamic-queue MCP server. See YAML format for the full list of group rules.