release-setup

Release Setup

Let: F := --force flag present in $ARGUMENTS σ := .claude/stack.yml D✅(label) := Display: {label} ✅ Configured D⏭(label) := Display: {label} ⏭ Already configured D⚠(label) := Display: {label} ⚠️ Install failed — check network/lockfile

Configure commit standards and release automation: Commitizen + commitlint, hook runner (Lefthook or Husky), and semantic-release or Release Please.

Can run standalone (/release-setup) or be called by /init after /ci-setup.

Phase 0 — Pre-check (idempotency)

Check prerequisites and per-component state before any installation.

1. Verify σ exists:

   test -f .claude/stack.yml && echo "found" || echo "missing"
   

   missing → warn: "stack.yml not found — release-setup reads runtime and hook runner from it." → DP(A) Run /env-setup first | Proceed manually Proceed manually → continue with defaults (runtime: node, package_manager: npm, hooks.tool: none).

2. Check per-component config file existence in parallel:

   test -f .lefthook.yml && echo "has_lefthook" || echo "no_lefthook"
   test -d .husky && echo "has_husky" || echo "no_husky"
   test -f .commitlintrc.cjs && echo "has_commits" || echo "no_commits"
   test -f release.config.cjs && echo "has_sr" || echo "no_sr"
   test -f release-please-config.json && echo "has_rp" || echo "no_rp"
   

3. Set booleans from results:

   • has_hook_runner := has_lefthook ∨ has_husky
   • has_commits := .commitlintrc.cjs ∃
   • has_releases := release.config.cjs ∃ ∨ release-please-config.json ∃
   • has_lefthook := .lefthook.yml ∃

4. F overrides all guards → treat all booleans as false (re-run all components).

Phase 1 — Stack Detection

Read configuration and detect environment.

1. Read σ fields: runtime, package_manager, hooks.tool, commands.lint, commands.typecheck. Defaults if σ missing: runtime=node, package_manager=npm, hooks.tool=none, commands.lint=npm run lint, commands.typecheck=npm run typecheck.

2. Detect existing hook runner via file checks:

   test -f .lefthook.yml && echo "lefthook" || (test -d .husky && echo "husky" || echo "none")
   

3. Detect branches:

   git branch -r | grep -E 'staging|develop|main|master'
   

   Store as branch_list. Default to ['main'] if only one or none detected.

Dispatch

Phase 2 — Hook Runner → Read ${CLAUDE_SKILL_DIR}/cookbooks/hook-runner.md, execute. Phase 3 — Commit Standards → Read ${CLAUDE_SKILL_DIR}/cookbooks/commit-standards.md, execute. Phase 4 — Release Automation → Read ${CLAUDE_SKILL_DIR}/cookbooks/release-automation.md, execute. Phase 5 — Summary (below).

Phase 5 — Summary (no auto-commit)

Display results and generated files. Do NOT run git add or git commit.

1. Display summary table:

   Release Setup Complete
   ======================
   
     Hook runner         ✅ Configured / ⏭ Already configured / ⏭ Skipped
     Commit standards    ✅ Configured / ⏭ Already configured / ⏭ Python not supported / ⏭ Skipped
     Release automation  ✅ Configured / ⏭ Already configured / ⏭ Skipped
   

2. List generated files (only those actually written):

   Generated files:
     .lefthook.yml
     .pre-commit-config.yaml       (Python only)
     .husky/pre-commit             (Husky only)
     .husky/commit-msg             (Husky + Commitizen)
     .commitlintrc.cjs             (Commitizen chosen)
     release.config.cjs            (semantic-release chosen)
     release-please-config.json    (Release Please chosen)
     .release-please-manifest.json (Release Please chosen)
   

3. Display suggested commit command:

   Suggested commit:
     git add <generated-files>
     git commit -m "chore: add release setup"
   

Safety Rules

1. Never push to remote without user confirmation
2. Always present decisions via protocol before installing packages or writing config files
3. Idempotent — skip already-configured components unless F
4. No auto-commit — sub-skill pattern: display files, let user review and commit

$ARGUMENTS