agents_guide

Guides delegation of focused work to child Agents with orchestration and tracking.

1.5k|172|Updated Mar 6, 2026
One-click install
npx skills add https://github.com/siddsachar/row-bot --skill agents-guide
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agents_guide
Source: https://github.com/siddsachar/row-bot/tree/main/tool_guides/agents_guide
Command: npx skills add https://github.com/siddsachar/row-bot --skill agents-guide

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Complex tasks often involve parallel workstreams, independent reviews, or long-running research that would clutter or overload a single conversation thread. This Skill provides the operational rules for delegating such work to scoped child Agents while keeping the parent thread responsive and accountable.

Core Features & Use Cases

  • Automatic Delegation Decisions: Defines when to spawn child Agents (parallel workstreams, bounded reviews, isolated writes) and when not to (trivial questions, sequential work).
  • Orchestration & Tracking: Covers delegate_work, agent_wait, agent_status, agent_message, agent_stop, and agent_retry for managing asynchronous child cohorts and joining results before finalizing.
  • Profiles & Workspaces: Explains how to select Agent Profiles, assign distinct developer workspace folders for parallel writers, and use git Worktrees for isolated file editing.
  • Use Case: When asked to research a topic and draft a report, delegate the research to a child Agent with the research profile, continue other work, then join the results with agent_wait before synthesizing the final answer.

Quick Start

Delegate the independent research and review tasks to child Agents, then wait for their results before giving me the final answer.

Frequently Asked Questions about agents_guide

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

FAQPage Schema
How do I delegate work to child Agents in Row-Bot?

Call delegate_work with a precise objective, a focused context packet, and optionally a profile such as research or review. Start ordinary automatic children with wait=false and required=true, then join results with agent_wait before finalizing.

When should I use child Agents instead of doing work directly?

Delegate when two or more substantial workstreams can proceed independently, a bounded review or research task protects parent context, or writes can be isolated in Worktrees. Avoid delegating trivial questions, single-step edits, or inherently sequential work.

Can child Agents run in parallel on the same files?

Children assigned to the same folder serialize due to workspace locking. For parallel writers, create separate existing folders first and pass a distinct developer_workspace_path to each delegate_work call, or use use_worktree=true for isolated git Worktrees.

How do I specify a different model for a child Agent?

The parent must pass a canonical pinned model ref to delegate_work(model=...) after inspecting active pinned choices with row_bot_status. Children cannot change their own model, and raw natural-language model phrases are not accepted.

What are the limitations of automatic Agent delegation?

Automatic delegation is unavailable when the Agents tool is disabled or the active profile forbids it. Children do not get recursive delegation unless the profile allows it, and depends_on only controls launch order without transferring outputs between children.