SuperAPI AI Directory

project-init

Agents & models

View repository

Updated 10-15 19:54

About this skill

Project Init Wizard

Run on explicit setup requests and deliberate context-recovery requests such as "Setup project", "Initialize context", "Generate CONTEXT.md", or "CONTEXT.md was deleted".

Core Rules

  1. Do not run blindly. Trigger only for setup intent or missing/incomplete context recovery intent.
  2. Ask one question at a time.
  3. Analyze local files before asking setup questions.
  4. Do not re-ask values that are already inferred with high confidence.
  5. If first user message is non-English, ask one language confirmation question before any other setup question.
  6. If .agents/CONTEXT.md has checklist and CONTEXT_READY is checked, skip analysis and only confirm whether updates are needed.
  7. If .agents/CONTEXT.md is missing but the repository is easy to infer, create a minimal recovery draft first and ask only for unresolved fields.
  8. Do not act as the default Codex entrypoint. Automatic Codex startup repair belongs to codex-precision.

Hard Trigger Policy

Trigger project-init with high priority when one of these is true:

  1. User explicitly asks setup/init/context generation.
  2. User says .agents/CONTEXT.md is missing, deleted, empty, outdated, or needs regeneration.
  3. User asks to remap skills from scratch for the project.

Do not stay only in global-config when the request matches this policy. Do not override codex-precision when the request is specifically about Codex startup self-heal, AGENTS.md tightening, or auto-fixing existing agent docs.

Workflow

Step 0: Conversation Language

Ask this only if first user message is non-English. This must be the first setup prompt.

O'zbekcha davom etamizmi?

1) Ha
2) Yo'q (English)
3) Boshqa til

Fallback template for other non-English languages:

Use {detected_language} for this setup conversation?

1) Yes
2) No (English)
3) Other language

If answer is 3, ask: Qaysi tilda davom etamiz? (or Which language? if user replies in English).

Step A: Quick Analysis and Confirmation

Scan and summarize:

  • .agents/CONTEXT.md
  • AGENTS.md
  • package.json
  • tsconfig.json
  • vite.config.*
  • next.config.*
  • nest-cli.json

Ask:

Here is the quick analysis summary. Continue with this direction or adjust it?

1) Continue
2) I want changes

If 2, apply user corrections first. If 1, continue.

Step A2: Minimal Self-Heal Draft

If enough can be inferred from repository files, create or repair .agents/CONTEXT.md immediately with:

  • detected stack
  • likely package manager
  • obvious commands from package.json
  • initial skill mapping
  • unchecked items left explicit in the checklist

Do not wait for a full wizard if the missing file can be restored to a usable draft safely.

Step B: Ask Only Missing Inputs

Collect only unresolved fields from this question bank:

Field Options
Project Name text
Project Type SaaS, E-commerce, Admin Dashboard, Landing Page, API Backend, Mobile App, Portfolio/Blog, Other
Frontend Framework Next.js, Vite + React, Remix, None
TypeScript Yes, No
UI Library Tailwind, shadcn + Tailwind, MUI, Ant Design, None
State Management TanStack Query, Zustand, Redux Toolkit, TanStack + Zustand, None
Backend NestJS, Express, Next.js API Routes, Hono, None
Database PostgreSQL, MySQL, MongoDB, SQLite, Supabase, None
ORM Prisma, Drizzle, TypeORM, None
Package Manager pnpm, npm, yarn, bun

Output Contract

Generate .agents/CONTEXT.md with this structure:

# {PROJECT_NAME} - Project Context

> Auto-generated by project-init wizard

## Context Checklist (for agents)
- [ ] CONTEXT_READY
- [ ] STACK_CONFIRMED
- [ ] STRUCTURE_CONFIRMED
- [ ] COMMANDS_CONFIRMED
- [ ] SKILLS_MAPPED
- [ ] NOTES_CONFIRMED
- [ ] LOCAL_SKILLS_SYNCED

## Tech Stack
| Category | Technology |
|----------|------------|
| Framework | {FRAMEWORK} |
| Language | {LANGUAGE} |
| Styling | {UI_LIBRARY} |
| State | {STATE_MANAGEMENT} |
| Backend | {BACKEND} |
| Database | {DATABASE} |
| ORM | {ORM} |
| Package Manager | {PACKAGE_MANAGER} |

## Skills
- Core: `global-config`, `errors`, `git`, `solid`
- Frontend: {FRONTEND_SKILLS}
- Backend: {BACKEND_SKILLS}
- Data: {DATA_SKILLS}
- State: {STATE_SKILLS}
- UI: {UI_SKILLS}
- Testing: {TESTING_SKILLS}

## Project Structure
{PROJECT_STRUCTURE}

## Coding Rules
- Components: Keep focused and small
- TypeScript: Strict mode, avoid `any`
- Commits: Conventional Commit format
- Naming: PascalCase components, camelCase utilities

## Commands
# pnpm
pnpm dev
pnpm build
pnpm lint
pnpm test

# npm
npm run dev
npm run build
npm run lint
npm run test

# yarn
yarn dev
yarn build
yarn lint
yarn test

# bun
bun run dev
bun run build
bun run lint
bun run test

Skill Mapping Rules

  • Frontend:
  • nextjs -> react-core, react-nextjs
  • vite -> react-core, react-vite
  • remix -> react-core
  • TypeScript:
  • yes -> typescript
  • UI:
  • tailwind -> tailwind
  • shadcn -> shadcn, tailwind
  • mui or ant-design -> design
  • State:
  • tanstack -> tanstack-query
  • zustand -> zustand
  • redux -> redux
  • Backend:
  • nestjs -> nestjs, api-patterns, security
  • express or next-api or hono -> api-patterns, security
  • Forms:
  • if required -> forms
  • Testing:
  • if planned -> testing
  • Any project:
  • always include git, solid, errors, global-config

References

  • For complete legacy question flow and examples: references/guide.md