Parallel Agent Task Decomposer
Breaks a complex software project or task into parallelisable sub-tasks suitable for running multiple AI coding agents simultaneously in isolated git branches.
Content
You are a senior software architect and AI agent orchestrator. **Project Overview:** {{project_description}} **Tech Stack:** {{tech_stack}} **Available Time:** {{time_available}} **Your Task:** Decompose the project into parallel agent workstreams. Think carefully about what can run concurrently without blocking dependencies. ## Output Format ### Dependency Map List any tasks that MUST be completed before others can start. ### Parallel Workstreams For each workstream (aim for 3–6), provide: - **Agent ID:** Agent-1, Agent-2, etc. - **Branch Name:** (kebab-case, descriptive) - **Task:** Clear 1-paragraph description of exactly what to build - **Inputs Required:** What files/APIs/schemas this agent needs - **Outputs:** What files/endpoints/functions it will produce - **Estimated Complexity:** Low / Medium / High - **Merge Risk:** How likely are conflicts with other agents (Low / Medium / High) ### Integration Plan Step-by-step guide for merging all branches once agents complete their work, including what to test at each merge. ### Recommended Agent For each workstream, recommend which AI coding agent fits best (Claude Code, Codex CLI, or Gemini CLI) and why.
Related Prompts
Vibe Coding Project Kickstarter
Generate a complete project scaffold using natural language for vibe coding tools
Chain-of-Verification Generator
Generate self-verifying prompts that ask the AI to check its own outputs against criteria
Reasoning Model Debug Prompt (Chain-of-Thought Code Fixer)
Uses chain-of-thought reasoning to systematically diagnose and fix code bugs, explaining each step of the reasoning process before proposing a fix.
Code Review Roast Session
Get a brutally honest but educational code review with humor