SuperAPI AI Directory

init

Coding & development

View repository

Updated 10-16 03:14

About this skill

Init

When to use

  • Starting a new repo from scratch
  • Turning a brief or spec into the first working scaffold
  • Applying your standard setup without restating the same preferences

Inputs

  1. Use $ARGUMENTS, the current brief, or the current spec as the project description.
  2. Read references/python-service.md for the built-in Python service defaults.
  3. If ~/.agents/stacks/python-service.md exists, read it and let it override the built-in Python reference.
  4. If the request is not a Python service or API, stop and ask for an explicit supported stack or a new stack reference.

Questions

Ask at most three concise questions in one message, and only if the answer is missing from the brief or spec:

  • project name
  • repo visibility, if creating a GitHub repo
  • any deviations from the stack defaults

Do not ask about standard tool choices already defined by the stack.

Process

  1. Read the brief or spec first. Infer project needs from the requirements instead of asking setup questions the stack already answers.
  2. Load the Python service reference and any user override.
  3. Infer conditional choices from the requirements:
    • If persistence is needed, use PostgreSQL.
    • If retrieval or RAG is needed, use PostgreSQL with pgvector.
    • If no persistence is implied, do not add database code.
  4. Check the latest stable versions of the core dependencies before pinning them. Do not rely on remembered versions.
  5. Generate the smallest working scaffold that follows the stack defaults.
  6. Write AGENTS.md with the standard just commands and repo rules so future agents inherit the same workflow.
  7. Run the strongest relevant verification before finishing.
  8. If asked, initialize git, make the first commit, and create the GitHub repo.

Output

Generate the standard project files, tool configuration, minimal app code, tests, and an AGENTS.md that tells future agents to use the just commands first.

Rules

  • Keep the scaffold lean. No empty folders, placeholder features, or speculative dependencies.
  • Prefer standards over invention.
  • Use FastAPI, pydantic-settings, uv, ruff, pyright, pytest, and just for Python services.
  • If a database is needed, default to PostgreSQL.
  • Document every required environment variable in .env.example.
  • Stop after scaffolding. Do not build product features.

Verification

  • just build
  • just test
  • just check
  • The app starts without import or settings errors
  • No secrets are committed