pivanov/split-tasks

Split Tasks Across Parallel Agents

Dispatch multiple Task agents simultaneously to complete independent work faster.

When to Use

• 3+ independent tasks with no shared dependencies
• Tasks that don't modify the same files
• Parallel investigations or searches

When NOT to Use

• Tasks depend on each other's output
• Multiple agents would edit the same file
• Sequential workflow required

Agent Types

| Type | Use for | |------|---------| | Bash | Running commands, tests, builds | | general-purpose | Code changes, file edits, research | | Explore | Codebase investigation, finding patterns |

Process

1. Identify remaining tasks from conversation
2. Filter to independent tasks only
3. Launch 3-6 Task agents in a SINGLE message (parallel tool calls)
4. Wait for all to complete
5. Consolidate results and report summary

Agent Prompt Template

Each agent needs:

• Goal: What to accomplish
• Context: Relevant background
• Success criteria: How to know it's done

Example

User: "Fix failing tests, update API docs, and add input validation to forms"

Launch in parallel:

• Agent 1 (Bash): Run tests, diagnose failures, fix them
• Agent 2 (general-purpose): Update API documentation to match current endpoints
• Agent 3 (general-purpose): Add validation to form components

After completion: "All 3 tasks complete. Tests passing (fixed X). Docs updated for Y endpoints. Validation added to Z forms."

Red Flags

• Never dispatch agents that touch the same files
• Never assume one agent's output for another (no dependencies)
• Never launch more than 6 agents (diminishing returns)
• Always consolidate results after all complete

If an Agent Fails

• Report which agent failed and why
• Do NOT retry automatically (let user decide)
• Other agents' work is still valid