workerflow-internal-workflow

Orchestrates temporary QwenPaw subagents and DAG workflows inside a single Worker.

5.5k|677|Updated Feb 21, 2026
One-click install
npx skills add https://github.com/agentscope-ai/HiClaw --skill workerflow-internal-workflow
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: workerflow-internal-workflow
Source: https://github.com/agentscope-ai/HiClaw/tree/main/plugins/workerflow/skills/agent/worker-internal-workflow
Command: npx skills add https://github.com/agentscope-ai/HiClaw --skill workerflow-internal-workflow

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

A single Worker often faces tasks too large for one pass but too small for full team delegation. This Skill guides a QwenPaw-backed Worker in choosing between doing work directly, using native subagents for short parallelism, or spinning up temporary QwenPaw agents with custom AGENTS.md templates and skills via the worker_agentflow workflow_run tool.

Core Features & Use Cases

  • Decision Framework: A decision table maps each need (direct work, native subagent, dynamic workflow, manual temp agent, team workflow) to the correct execution path.
  • Dynamic Workflow Lifecycle: Covers workflow_run, workflow_update, workflow_finish, and workflow_fail, including DAG nodes with dependsOn dependencies and readyInstructions scheduling.
  • Shared File System: Defines a run-level shared directory with inputs/ and per-agent outputs/ paths so temporary agents exchange files safely.
  • Use Case: A Worker reviewing an auth module fans out to security-reviewer and test-reviewer subagents, merges their fixed-shape findings, deletes the temporary agents, and reports one severity-ordered list.

Quick Start

Ask the Worker to split the current task into parallel subagent reviews using workerflow and merge the results into one report.

Frequently Asked Questions about workerflow-internal-workflow

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I split work inside a single QwenPaw Worker?

Choose based on need: do it directly for simple tasks, use native subagents for short parallel inspection, or call worker_agentflow workflow_run when you need custom AGENTS.md templates, separate workspaces, or visible fan-out. Persistent team members require the team workflow instead.

How do I run a DAG workflow with dependent subagents?

Define nodes with dependsOn arrays in workflow_run. Only dependency-free nodes appear in submitInstructions; after each subagent finishes, call workflow_update with done steps and submit any newly returned readyInstructions to downstream agents.

When should I use a temporary agent versus a native subagent?

Use native subagents for short internal parallelism sharing the current workspace. Create a temporary agent only when you need a custom prompt, workspace, or skill set, and always delete it with delete_temp_agent after the bounded task completes or fails.

How do temporary agents share files in workerflow?

workflow_run and create_temp_agent expose a shared directory at <default-workspace>/shared/workerflow/<sharedRunId>/. Place inputs under shared/inputs/ and have each subagent write only under shared/outputs/<agent-id>/, passing absolute paths in prompts.

Why must temporary agents use tmp- prefixed ids?

Temporary agents are implementation details of the current Worker, so tmp- ids distinguish them from durable team members. They must never be registered as TeamHarness workers and must be deleted as a finally step after completion or failure.