agentleFS
glebisglebis/claude-skills/init-tauri-app/SKILL.md

init-tauri-app

Skill400 starsChanged 4 months agoMITView on GitHub

Scaffold a new Tauri v2 project with the cenno/cull house conventions — delegates boilerplate to `npm create tauri-app`, then layers an opinionated core plus opt-in modules (CLI+MCP, SQLite, tray/updater, release/preflight, Swift sidecar). Use when the user wants to start a new Tauri desktop app, "init a tauri project", or "scaffold a tauri app".

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Ask, in one batch:

    line 23
  2. Very specific:

    If Swift sidecar selected but host is non-macOS or xcrun --find swift fails: warn and drop it.

    line 32
  3. Distinctive:

    If none found, skip this whole step (the scaffold proceeds with empty product context — no error).

    line 36
  4. Very specific:

    Pre-fill: if valid, default the app name to name and identifier to com.glebkalinin.<name> (still confirm with the user in Step 1 if not …

    line 40
  5. Very specific:

    Create empty tracked dir docs/internal/.gitkeep and docs/.gitkeep.

    line 83

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

init-tauri-app

Scaffolds a new Tauri v2 project pre-loaded with conventions proven in two production apps (cenno, cull). Delegates version-current boilerplate to the official scaffolder, then applies a durable convention layer and any opt-in modules the user selects.

When to use

  • "Start a new Tauri app", "init a tauri project", "scaffold a tauri desktop app".

Prerequisites (verify before scaffolding)

  • node + npm on PATH (node --version)
  • cargo + rustc on PATH (cargo --version)
  • macOS + Xcode CLT only required if the Swift-sidecar module is selected

Procedure

1. Gather inputs (AskUserQuestion)

Ask, in one batch:

  • App name (kebab-case). Validate ^[a-z][a-z0-9-]*$.
  • Identifier (reverse-DNS, default com.glebkalinin.<name>).
  • Target directory (default ~/ai_projects/<name>). Abort if it exists and is non-empty.
  • Frontend framework: react-ts | svelte-kit | vanilla-ts.
  • Modules (multi-select): CLI+MCP · SQLite · Tray/Updater · Release/Preflight · Swift sidecar.
  • JTBD artifact (optional): a path to a jtbd.json. If not given, the skill auto-discovers ./jtbd.json then ~/jtbd/<name>/jtbd.json. See Step 1.5.

If Swift sidecar selected but host is non-macOS or xcrun --find swift fails: warn and drop it.

1.5 Ingest JTBD (optional, additive)

  1. Resolve the artifact, first hit wins: explicit path → ./jtbd.json → ~/jtbd/<name>/jtbd.json. If none found, skip this whole step (the scaffold proceeds with empty product context — no error).
  2. Confirm: echo the artifact's hook and ask the user to confirm before using it. On decline, skip.
  3. Validate: the artifact must parse and have name, hook, jtbd. If not, warn and skip ingestion (never abort the scaffold). render-jtbd.sh exits 3 on invalid input — treat that as "skip".
  4. Pre-fill: if valid, default the app name to name and identifier to com.glebkalinin.<name> (still confirm with the user in Step 1 if not already chosen).
  5. The artifacts are written during Step 3 (core layer) — see the JTBD block there.

2. Scaffold base

cd <parent-of-target>
npm create tauri-app@latest <name> -- --template <framework> --manager npm --yes

Then cd <target> && npm install.

Known upstream fix (current rustc + Tauri 2.11.x): a bare scaffold's first cargo check can fail with error[E0119] on cookie/time (transitive cookie 0.18.1 vs time 0.3.48). This is not a skill bug — it hits any fresh create-tauri-app. If it occurs, run once in src-tauri/: cargo update -p time --precise 0.3.47, then re-check.

3. Apply core layer

Copy every file from assets/core/ into the project, applying the renames in the table below, then run a baseline gate. Substitute <name> (and <identifier> where the token appears) in AGENTS.md and README.

assetdestination
core/AGENTS.mdAGENTS.md
core/CLAUDE.md.claude/CLAUDE.md
core/gitignore.gitignore (merge: append only house lines not already present; skip lines — including comment headers — that already exist)
core/rust-toolchain.tomlsrc-tauri/rust-toolchain.toml
core/node-version.node-version
core/mcp.json.mcp.json
core/capabilities/default.jsonsrc-tauri/capabilities/default.json (overwrite)
core/scripts/check-versions.shscripts/check-versions.sh (chmod +x)
core/README.mdREADME.md
core/CONTRIBUTING.mdCONTRIBUTING.md

If a JTBD artifact was confirmed in Step 1.5, also:

  • Render assets/jtbd/PRODUCT.md.template → docs/PRODUCT.md via scripts/render-jtbd.sh <artifact> assets/jtbd/PRODUCT.md.template <artifact-path>.
  • Render assets/jtbd/guardrails-check.md.template → docs/internal/guardrails-check.md.
  • Render assets/jtbd/agents-product-section.md.template and insert it into AGENTS.md immediately after the first heading (so product context leads the file).
  • Copy the artifact verbatim to project-root jtbd.json (never modify the source).
  • Field→destination details: assets/jtbd/jtbd-map.md.

Create empty tracked dir docs/internal/.gitkeep and docs/.gitkeep. Enable TS strict: ensure tsconfig.json has strict, noUnusedLocals, noUnusedParameters true.

Gate: cd src-tauri && cargo check and cd .. && npm run build. Both must pass before modules.

4. Compose selected modules

For each selected module, in this order — cli-mcp, sqlite, tray-updater, swift-sidecar, release-preflight — open assets/modules/<m>/INSERT.md and follow it exactly: it lists files to copy, Cargo deps to merge into src-tauri/Cargo.toml [dependencies], and insertion points in src-tauri/src/lib.rs (tauri::generate_handler![...]) and src-tauri/src/main.rs.

After EACH module: cd src-tauri && cargo check. If it fails, fix the just-applied merge before continuing (failures localize to the current module). For modules with a frontend/script part, also run the relevant check named in that INSERT.md.

5. Final verification + handoff

  • cd src-tauri && cargo check → must pass
  • npm run build → must pass
  • bash scripts/check-versions.sh → must pass
  • Offer git init && git add -A && git commit -m "chore: scaffold via init-tauri-app".
  • Print a summary: framework, modules applied, modules skipped (with reason), next commands (npm run tauri dev).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.