Turn a raw idea into a practical execution plan for this repository.
Core Outcome
Produce two files at project root:
- •
docs/ideation/PLAN.md: phase-by-phase implementation plan. - •
docs/ideation/tasks.json: machine-readable task list grouped by phase.
Use this repository's architecture and commands, not generic templates.
Required Workflow
- •Confirm scope from user input
- •Extract goal, target users, core workflow, and non-goals.
- •Record assumptions if input is incomplete.
- •Ask only required clarifying questions
- •Ask questions only when missing answers materially change architecture or phase order.
- •Ask at most 5 concise questions in one round.
- •Prefer closed choices when possible.
- •If user does not answer, continue with explicit assumptions.
- •Load repository context
- •Read
AGENTS.md,README.md, rootpackage.json,frontend/package.json, andserver/package.json. - •Align planning to:
- •
frontend/for UI, pages, state, API clients, and tests. - •
server/for Hono routes, middleware, services, and tests. - •
shared/for shared types and config.
- •
- •Perform focused research when needed
- •Do local repo research first.
- •Do web research if the idea depends on unfamiliar libraries, third-party APIs, security/compliance requirements, or complex domain rules.
- •Use context7 MCP server to read docs of frameworks if needed
- •Capture findings as short bullets with links inside
PLAN.mdunderResearch Notes. - •Distinguish confirmed facts from assumptions.
- •Design phased implementation
- •Break work into ordered phases where each phase can be validated.
- •Keep phases outcome-driven.
- •Define dependencies and risk per phase.
- •Ensure each phase maps to concrete folders/files in this repo.
- •Generate
PLAN.md
- •Use
references/plan-template.md. - •Keep language specific and implementation-focused.
- •Include architecture decisions for frontend, server, shared, and Firebase usage.
- •Include testing, observability, and rollout strategy.
- •Include testing tasks for both happy path and key error paths.
- •Generate
tasks.json
- •This is an array of tasks that a user or AI needs to perform step by step in order to complete a plan phase
- •Use template from
tasks-template.json - •Convert each phase into actionable task objects.
- •Include acceptance criteria and dependencies per task.
- •Run quality gate before finalizing
- •Verify
PLAN.mdandtasks.jsonare consistent and phase names match exactly. - •Ensure no phase is missing test tasks.
- •Ensure commands use Bun and existing workspace scripts.
- •Ensure unknowns are captured in
Open Questionsrather than hidden.
Planning Rules
- •Prefer simple vertical slices before broad platform work.
- •Plan MVP-first; defer optional enhancements to later phases.
- •Keep each phase independently reviewable.
- •Do not include code unless user explicitly asks.
- •Do not invent APIs, environment variables, or infra details without labeling assumptions.
Output Contract
PLAN.md must include:
- •Problem Statement
- •Solution Overview
- •Assumptions and Constraints
- •Repo Impact Map
- •Phase Plan
- •Testing Strategy
- •Risks and Mitigations
- •Research Notes
- •Open Questions
tasks.json must include:
- •Top-level metadata:
idea,generatedAt,repo,phases - •Phase entries matching
PLAN.mdnames - •Task objects with required fields:
- •
id(number) - •
status(pending, done) - •
title(task title) - •
description(detailed task context) - •
notes(any additional notes from previous tasks)
- •
Task object example:
json
{
"id": "task-001",
"status": "todo",
"title": "Create todo route",
"description": "Add Hono endpoints for todo CRUD with validation and error handling.",
"notes": ["Endpoints should be under /api/"]
}
Load these references when needed:
- •
plan-template.md - •
tasks-template.json