takazudo/b4push-wisdom

B4Push Wisdom — Full Validation Setup Guide

Set up comprehensive before-push validation and CI for any project. This covers three parts:

1. b4push script — local validation before pushing
2. b4push skill — Claude Code skill to run b4push with auto-fix
3. CI workflow — GitHub Actions to enforce checks on PRs and main

Step 1: Analyze the project

Read package.json and explore the project structure to understand:

• Package manager: Check for pnpm-lock.yaml, package-lock.json, yarn.lock
• Available scripts: check, build, test, lint, format, typecheck
• Workspace packages: Check pnpm-workspace.yaml for sub-packages with their own tests
• Doc site: Look for doc/, docs/, website/ directories with their own package.json
• E2E tests: Check for playwright.config.*, cypress.config.*
• Data generation: Check for generate-* scripts

Step 2: Determine b4push steps

Common patterns (adapt to project):

| Step | Command | When to include | | --- | --- | --- | | Workspace tests | pnpm --filter "@scope/*" test | If pnpm workspace with test scripts | | App unit tests | pnpm --filter app-name test:unit | If app has unit tests | | Code quality | pnpm check or pnpm lint && pnpm format | Always | | TypeScript | pnpm typecheck or pnpm --filter name typecheck | If TypeScript | | Build | pnpm build | If build script exists | | Doc quality | cd doc && pnpm check | If doc site has check script | | Doc build | cd doc && pnpm build | If doc site exists | | E2E tests | Start server + run playwright | If e2e tests exist |

Step 3: Create scripts/run-b4push.sh

Use this template structure:

#!/usr/bin/env bash
set -euo pipefail

START_TIME=$(date +%s)
FAILURES=()

step() {
  echo ""
  echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
  echo "▶ $1"
  echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
}

pass() { echo "✅ $1"; }
fail() { echo "❌ $1"; FAILURES+=("$1"); }

ROOT_DIR="$(cd "$(dirname "$0")/.." && pwd)"

# Steps here — each wrapped in:
# step "Step N/M: Description"
# if (cd "$ROOT_DIR" && command); then
#   pass "Description passed"
# else
#   fail "Description"
# fi

# Summary
END_TIME=$(date +%s)
DURATION=$((END_TIME - START_TIME))

echo ""
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
echo "  SUMMARY (${DURATION}s)"
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"

if [ ${#FAILURES[@]} -eq 0 ]; then
  echo "✅ All checks passed! Safe to push."
  exit 0
else
  echo "❌ ${#FAILURES[@]} check(s) failed:"
  for f in "${FAILURES[@]}"; do
    echo "   - $f"
  done
  exit 1
fi

Key rules:

• set -euo pipefail for strict error handling
• Continue all steps even if some fail (collect in FAILURES array)
• Subshell execution (cd "$ROOT_DIR" && command) to isolate
• Summary at the end with elapsed time

Make executable: chmod +x scripts/run-b4push.sh

Step 4: Add package.json script

{
  "scripts": {
    "b4push": "./scripts/run-b4push.sh"
  }
}

Step 5: Create project-specific b4push skill

Create .claude/skills/b4push/SKILL.md (uppercase — the canonical name; lowercase skill.md causes git dual-tracking / clone collisions on case-insensitive filesystems):

---
name: b4push
description: >-
  Run comprehensive pre-push validation covering [list steps]. Use when: (1) Completing a PR
  or feature implementation, (2) Before pushing significant changes, (3) After large refactors,
  (4) User says 'b4push', 'before push', 'check everything', or 'ready to push'.
user-invocable: true
allowed-tools:
  - Bash
---

# Before Push Check

Run `pnpm b4push` from the project root. This executes `scripts/run-b4push.sh`:

1. [Step list with descriptions]

Takes ~[duration]. All steps must pass.

## On failure

1. Read the failure output to identify which step failed
2. Auto-fix what you can:
   - Formatting: `pnpm check:fix` or `cd doc && pnpm check:fix`
   - Lint: `pnpm lint:fix` or `cd doc && pnpm lint:fix`
3. Re-run `pnpm b4push` to confirm all checks pass
4. Report the final status

Step 6: Create GitHub Actions CI workflow (optional)

If the project uses GitHub and doesn't have CI yet, create .github/workflows/ci.yml:

name: CI

on:
  pull_request:
    types: [opened, synchronize, reopened]
    branches: [main]
  push:
    branches: [main]

concurrency:
  group: ci-${{ github.head_ref || github.ref }}
  cancel-in-progress: true

jobs:
  checks:
    name: Checks
    runs-on: ubuntu-latest
    timeout-minutes: 15

    steps:
      - uses: actions/checkout@v4

      - uses: pnpm/action-setup@v4
        with:
          version: 10

      - uses: actions/setup-node@v4
        with:
          node-version: '20'

      - run: pnpm install --frozen-lockfile

      # Mirror the b4push steps here
      - name: [Step description]
        run: [command]

The CI workflow should mirror the b4push steps so local and CI validation are consistent.

Key CI patterns:

• Cancel previous runs for same PR (concurrency group)
• Use pnpm/action-setup@v4 + actions/setup-node@v4 with no cache: (fresh install from the npm CDN beats cache restore — see gh-actions-wisdom rule 5)
• pnpm install --frozen-lockfile for reproducible installs
• Each step as a separate run for clear failure identification

Step 7: Test

Run pnpm b4push to verify all steps execute correctly. Fix any issues found.

Reference projects

These projects have working b4push setups:

• message (zudomessages): 4 steps (workspace tests, app tests, doc checks, doc build), ~1-2 min
• mdx-formatter: 6 steps (quality, build, test, doc data, doc quality, doc build), ~40s
• zmod: 9 steps including e2e with production server, ~3-4 min
• zpanels: Dual-track (quick ~3-5 min, full with e2e ~10-15 min)