Spawn and manage autonomous background agents for parallel task execution.
npx skills add https://github.com/Logarn/Worklin-ai --skill subagent-logarn
Please help me install this Agent Skill. Skill: subagent Source: https://github.com/Logarn/Worklin-ai/tree/main/assistant/src/config/bundled-skills/subagent Command: npx skills add https://github.com/Logarn/Worklin-ai --skill subagent-logarn
subagent_spawn. - Lifecycle Management: Track, message, read output from, and abort running subagents using subagent_status, subagent_message, subagent_read, and subagent_abort. - Fork Mode: Create context-inheriting forks that share the parent's full conversation and memory for tasks that depend on prior discussion. - Use Case: While investigating a production bug, spawn an investigator subagent to trace logs and a researcher to check related documentation in parallel, then read their compact reports when they complete. ## Quick Start Spawn a researcher subagent labeled "research-auth-libraries" to compare authentication options and report back when done.High-intent search queries and answers about installing and using this skill.
Use subagent_spawn with a label, objective, and role for each task. Each subagent runs autonomously in the background, and the parent conversation is automatically notified when each one completes, fails, or is aborted.
A fork inherits the parent's full context including messages, system prompt, and memory, while a regular subagent receives only its objective and context fields. Forks always run as the general role and default to internal-only results.
Use the investigator role for root-cause analysis, debugging, and log forensics. It has read-only shell access for grep and log reading plus file and web tools, and returns a compact root-cause report instead of flooding the main context.
Only subagents with the supervisor role can spawn one level of child workers. Delegation is bounded to depth 2, at most 8 active children per parent, and at most 16 total descendants across the tree.
Avoid spawning for tasks small enough to do inline, such as a single tool call, one search, or a quick file read. Subagents are intended for web research, multi-file exploration, and independent coding work that benefits from parallel execution.
VoTruongDanhDecompose complex requests into coordinated execution plans with agent roles and dependencies.
artosetrovOrchestrate multi-agent workflows into phased execution plans with agent-bus communication.
RogerGimbelSearch agent memory and session history using QMD semantic search.
CUexterDelegate coding tasks to Claude Code via the Hermes terminal.