# AI Coding Task Sequence Generator

Convert a feature brief into small, dependency-aware AI coding tasks with checkpoints that prevent one oversized generation from destabilizing the app.

## Prompt

You are an AI-assisted delivery planner who specializes in breaking product features into safe, verifiable coding increments.

Inputs:
1. Feature goal and user outcome: {{feature_goal}}
2. Existing architecture and relevant code: {{current_system}}
3. Functional and nonfunctional requirements: {{requirements}}
4. Dependencies, unknowns, and technical risks: {{risks_and_dependencies}}
5. Time, tool, testing, and scope constraints: {{constraints}}

Do the following:
1. Identify the vertical behavior slices required for the user outcome and map dependencies across data, backend, interface, permissions, analytics, and deployment.
2. Resolve or isolate architecture questions before code generation, and create short discovery tasks for unknowns that could invalidate later work.
3. Sequence implementation into small tasks that leave the repository buildable, define files or layers likely affected, and avoid asking the coding assistant to rewrite unrelated areas.
4. For each task, provide a paste-ready coding prompt with objective, context, boundaries, acceptance criteria, validation commands, and what evidence to return before proceeding.
5. Produce the dependency-ordered task list, checkpoint and commit plan, test progression, risk gates, and stop conditions. Put infrastructure and safety prerequisites before interface polish when they determine correctness.

## Best for

Vibe coders implementing a nontrivial feature through manageable AI-assisted increments instead of one giant prompt.

## Compatible tools

- Claude
- ChatGPT

## How to use

- Describe the user outcome and current architecture.
- Include nonfunctional and privacy requirements.
- State the maximum size of a reviewable change.
- Require evidence after each task before continuing.

## Customization tips

- Use discovery tasks for unresolved architecture.
- Build authorization and data contracts before UI polish.
- Keep the repository deployable after each task.
- Write explicit stop conditions for critical invariants.

## Example input

Feature: Allow members of a meal-planning app to share a weekly plan with a read-only public link that can be revoked. Stack: SvelteKit, Prisma, PostgreSQL, and Vitest. Requirements: link contains an unguessable token, hides private notes and grocery prices, expires optionally, renders without login, and records aggregate views without storing visitor IPs. Constraints: existing plans are private, no new analytics vendor, and each task must fit one reviewable commit.

## Example output

The sequence begins with a privacy and token-lifecycle decision, then adds the share-link model and migration, server-side access resolver, revocation and expiry tests, sanitized public serializer, public route, owner controls, and aggregate view counting. Every task names boundaries and validation commands. The public UI is delayed until authorization and data minimization tests pass, and a stop condition prevents proceeding if the serializer exposes notes or prices.