SuperAPI AI Directory

setup

General & other

View repository

Updated 10-16 00:21

About this skill

Project setup

Make the next feature easy to locate, run, and verify. Inspect the repository, manifests, lockfile, and local instructions first. Improve existing conventions before introducing new ones.

Work

  • Confirm runtime, package manager, framework, deploy target, and actual commands. Distinguish existing tooling from proposed additions.
  • Place route composition, feature logic, reusable primitives, validation, and data access near their owners. Do not create empty folders for hypothetical scale.
  • Choose one token source and identify how primitives consume it. Reuse an existing UI library; a new repository does not require a custom Button.
  • Document environment variable names and safe placeholders. Keep secret validation and access server-side; never import a secret-bearing environment object into browser code.
  • Set the smallest useful verification path: types/lint/build and a meaningful check of the first critical flow. Add infrastructure when it addresses a concrete risk.
  • On existing projects, preserve user edits and change only the foundation implicated by the task. Explain migration effects on consumers.

Deliver

For planning, give actual/proposed paths, commands, and an ordered setup sequence. For implementation, create the authorized foundation and run its checks rather than stopping at a checklist.

## Foundation plan

Stack assumption: Package manager: Deploy target: Structure: actual/proposed paths Commands: install, dev, check, test Env: names + placeholders + validation Tokens / data / checks: locations and commands

Setup sequence

  1. ...
  2. ...

A usable result lets a fresh checkout install reproducibly, run a slice, and locate tokens, data access, and verification commands. Update the package-manager lockfile through the package manager when dependencies change.

Use `$context` to record durable repository facts and `$git` for collaboration policy when those are needed.

## Worked example

```markdown
## Foundation plan

**Stack assumption:** Detected Vue 3 + Vite from manifest; preserved
**Package manager:** Detected npm from lockfile; preserved
**Deploy target:** Detected host from repo config; preserved
**Structure:** Detected `app/`, `ui/`, and `lib/` folders preserved; the first feature gets its own folder once it has files
**Commands:** Detected install, dev, check, test from manifest; no new tooling added
**Env:** `.env.example` added for the three variables the code reads; server-side validation module; secrets stay server-side
**Tokens / data / checks:** Reuse detected token file; data access stays in `lib/`; detected check runs before PRs

### Setup sequence
1. Record detected stack and commands
2. Repair missing env example
3. Reuse token source
4. Build first slice + check
5. Record facts with `$context`

Gotchas

  • The lockfile, not the README, names the package manager; installing with another one creates a second lockfile.
  • Many frameworks ship variables with a public prefix, and any module a client file imports, to the browser; check how this project keeps secrets server-side.
  • Empty folders for future scale (services/, utils/) become dumping grounds; create a folder with its first file.
  • A setup that works on one machine often leans on a global tool or a local variable; test from a fresh clone with only the documented steps.
  • A foundation repair that moves files breaks every open branch; list the moves for other contributors.

Boundaries

  • Do not use to choose the framework or stack; use $stack instead.
  • Do not use to audit UI drift or duplicate components; use $system instead.