Parallel Agent Execution Skill

This skill enables efficient multi-agent coordination for complex tasks.

When to Use

• Complex tasks requiring multiple exploration paths
• Research + implementation in parallel
• Large refactoring across multiple files
• Tasks benefiting from diverse approaches

Available Agents

Built-in Agents

Agent	Purpose	Mode
@agent	Autonomous coding	Full access
#ask	Q&A, explanations	Read-only
#edit	Direct file editing	Write
Plan	Research & planning	Read-only
Research	KB lookup	Read-only
Planner	Implementation design	Read-only

GitHub Copilot Coding Agent

For async background work that creates PRs:

Use: mcp_github_github_assign_copilot_to_issue
Result: Creates branch (copilot/*) and opens PR

Parallelism Patterns

Pattern 1: Explore-Then-Implement

1. Fire Research agent for solutions
2. Fire Planner agent for design
3. Wait for both results
4. Implement using @agent

Pattern 2: Multi-File Refactor

1. Use runSubagent for each file group
2. Coordinate changes via todo list
3. Apply changes sequentially

Pattern 3: Research + Code

1. Parallel: KB search + web fetch + file read
2. Sequential: Plan → Implement → Test

Best Practices

1. Never parallelize terminal commands - Run sequentially
2. Parallelize read operations - file reads, searches, fetches
3. Use todo list - Track multi-step progress
4. Delegate exploration - Use subagents for research
5. Commit per turn - When using background agents

Task Delegation Table

Task Type	Recommended Agent
Algorithm research	Research subagent
Code generation	@agent
File exploration	runSubagent
Quick edits	#edit
Planning	Plan subagent
Async PR work	Copilot Coding Agent