Build ideas from scratch through guided discovery, then produce developer-ready specs and implementation prompts.
- Starting a new side project from a rough idea
- Need structured thinking before jumping into code
- Want implementation prompts for a code-generation LLM
We're going to flesh out an idea together. Ask me ~20-25 questions, one at a time, building on your answers until we have enough detail for a solid specification.
You must:
- Challenge assumptions to ensure the idea meets real user needs
- Dig into edge cases and potential issues
- Keep going until we've covered all relevant aspects
- Remember: ask one question at a time
Paste your idea below:
---IDEA---
Compile our brainstorming into a comprehensive, developer-ready specification using the Cloudflare Workers platform where viable.
Output Structure:
- Overview — Problem, solution, target users
- Requirements — Functional and non-functional
- Architecture — Components, data flow, Cloudflare services used
- Data Model — Schemas, storage (KV, D1, R2, etc.)
- API Design — Endpoints, request/response formats
- Security — Authentication, authorization, and cost protection
- Error Handling — Failure modes and recovery strategies
- Testing Plan — Unit, integration, and edge case coverage
Security Section Must Include:
- Routes that trigger paid API calls (OpenAI, etc.) and their protection strategy
- Authentication requirements per route (public vs. authenticated)
- Fail-closed vs. fail-open behavior for auth middleware
- Rate limiting strategy to prevent abuse
- Any debug/development routes and their production disposition
Turn the spec into a series of implementation prompts for a code-generation LLM.
Process:
- Draft a step-by-step implementation blueprint
- Break into small, iterative chunks that build on each other
- Review and re-chunk until steps are right-sized:
- Small enough for safe implementation with strong testing
- Large enough to make meaningful progress
- Generate a prompt for each step
Prompt Requirements:
- Each prompt builds on previous work
- No orphaned code — everything integrates immediately
- Prioritize: performance, best practices, early testing
- No large complexity jumps between steps
Output Format: Each prompt in a fenced code block, with brief context above explaining what it accomplishes.
Create a new .md file in the same directory as the spec. Name it prompt-plan.md.
Spec:
---INSERT SPEC---
Generate todo.md — a markdown checklist covering the full project (spec + prompt plan). Be thorough.