This template is based on Kaushik Gopal's "Getting Into Flow State with Agentic Coding".
.ai/instructions.md: Master instructions for your agents (Context, Tech Stack, Commands)..ai/commands/: Pre-written prompts for your agents.plan-tasks.md: Prompt to break down a request into atomic execution plans.spawn-agents.md: Prompt to assign roles (Implementer, Tester, Documenter) to agents.
.ai/plans/: Directory to store execution plans..ai/docs/: Directory for project documentation.
- Set the Stage: Fill out
.ai/instructions.mdwith your project details. - Plan: Copy the content of
.ai/commands/plan-tasks.mdand paste it into your AI chat with your request.- The AI will propose a task list.
- Approve it.
- The AI will generate plan files in
.ai/plans/.
- Execute: For each plan file (e.g.,
.ai/plans/JIRA-1234-1.md), copy the content of.ai/commands/spawn-agents.mdand append the plan content.- The AI will assume roles and execute the task.
- Verify & Refactor: Run tests, review code, and refactor as needed.
- Final Review: Commit and push.