SKILL: GitHub Actions Testing & Validation Expert

Description

Interactive expert for testing and validating GitHub Actions workflows before deployment. Prevents common CI failures by catching cache configuration errors, path issues, monorepo dependency problems, and service container configuration mistakes.

Capabilities

This skill provides:

• Pre-Push Validation: Complete workflow validation before pushing to GitHub

• Cache Configuration: Ensure cache-dependency-path is correctly specified

• Monorepo Build Order: Validate workspace dependency build sequences

• Service Container Setup: Guide proper service container configuration

• Path Validation: Verify all paths exist and are accessible

• Local Testing: Run workflows locally with act (Docker-based simulation)

• Static Analysis: Lint workflows with actionlint and yamllint

When to Use This Skill

Invoke this skill when:

• Creating or modifying GitHub Actions workflows

• Debugging workflow failures in CI

• Setting up new repositories with CI/CD

• Migrating to monorepo architecture

• Adding service containers to workflows

• Experiencing cache-related failures

• Getting "module not found" errors in CI but not locally

Usage

Quick Validation

"Validate my GitHub Actions workflows before I push"

I'll:

• Run actionlint on all workflow files

• Check for missing cache-dependency-path configurations

• Validate all working-directory paths exist

• Verify monorepo build order is correct

• Check service container configurations

• Provide a pre-push checklist

Debugging Workflow Failures

"My GitHub Actions workflow is failing with [error message]"

I'll:

• Analyze the error message

• Identify the root cause

• Explain why local testing didn't catch it

• Provide the correct configuration

• Show how to test the fix locally

Setup New Repository

"Set up GitHub Actions testing for my new project"

I'll:

• Install required tools (act, actionlint, yamllint)

• Create validation scripts

• Set up pre-push hooks

• Configure recommended workflows

• Provide testing procedures

Critical Rules I Enforce

1. Cache Configuration

ALWAYS specify cache-dependency-path explicitly:

❌ WRONG

• uses: actions/setup-node@v4 with: cache: 'npm'

✅ CORRECT

• uses: actions/setup-node@v4 with: cache: 'npm' cache-dependency-path: package-lock.json

Why: GitHub Actions cache resolution fails silently in local testing but errors in CI with "Some specified paths were not resolved, unable to cache dependencies."

2. Monorepo Build Order

ALWAYS build workspace dependencies before type checking:

❌ WRONG

• run: npm ci
• run: npx tsc --noEmit

✅ CORRECT

• run: npm ci
• run: npm run build --workspace=@prpm/types
• run: npm run build --workspace=@prpm/registry-client
• run: npx tsc --noEmit

Why: TypeScript needs compiled output from workspace dependencies. Local development has pre-built artifacts, but CI starts clean.

3. npm ci in Monorepos

ALWAYS run npm ci from root, not workspace directories:

❌ WRONG

• working-directory: packages/infra run: npm ci

✅ CORRECT

• run: npm ci
• working-directory: packages/infra run: pulumi preview

Why: npm workspaces are managed from root. Workspace directories don't have their own package-lock.json.

4. Service Containers

Service containers can't override CMD via options:

❌ WRONG

services: minio: image: minio/minio:latest options: server /data # Ignored!

✅ CORRECT

services: minio: image: minio/minio:latest

steps:

• run: | docker exec $(docker ps -q --filter ancestor=minio/minio:latest)
  sh -c "minio server /data &"

Why: GitHub Actions service containers ignore custom commands. They must be started manually in steps.

Validation Tools

Required Tools

macOS

brew install act actionlint yamllint

Linux

curl https://raw.githubusercontent.com/nektos/act/master/install.sh | sudo bash bash <(curl https://raw.githubusercontent.com/rhysd/actionlint/main/scripts/download-actionlint.bash) pip install yamllint

Validation Script

I'll create .github/scripts/validate-workflows.sh :

#!/bin/bash set -e

echo "🔍 Validating GitHub Actions workflows..."

1. Static analysis

actionlint .github/workflows/.yml yamllint .github/workflows/.yml

2. Cache configuration check

for file in .github/workflows/*.yml; do if grep -q "cache: 'npm'" "$file"; then if ! grep -A 2 "cache: 'npm'" "$file" | grep -q "cache-dependency-path"; then echo "❌ $file: Missing explicit cache-dependency-path" exit 1 fi fi done

3. Path validation

grep -r "working-directory:" .github/workflows/*.yml | while read -r line; do dir=$(echo "$line" | sed 's/.*working-directory: //' | tr -d '"') if [ ! -d "$dir" ]; then echo "❌ Directory does not exist: $dir" exit 1 fi done

4. Check for explicit cache paths

grep -r "cache-dependency-path:" .github/workflows/*.yml | while read -r line; do path=$(echo "$line" | sed 's/.*cache-dependency-path: //' | tr -d '"') if [ ! -f "$path" ]; then echo "❌ Cache dependency path does not exist: $path" exit 1 fi done

echo "✅ All workflow validations passed"

Pre-Push Checklist

Before pushing workflow changes:

• Lint: actionlint .github/workflows/*.yml

• Validate: .github/scripts/validate-workflows.sh

• Dry Run: act pull_request -W .github/workflows/[workflow].yml -n

• Check Cache Paths: Verify all cache-dependency-path values exist

• Check Build Order: Ensure workspace dependencies built before type checks

• Service Containers: Confirm manual startup if custom commands needed

Common Failure Patterns

"Cannot find module '@prpm/types'"

Root Cause: Workspace dependency not built before type checking

Why Local Works: Previous builds exist in node_modules/

Fix:

• name: Build @prpm/types run: npm run build --workspace=@prpm/types
• name: Type check run: npx tsc --noEmit

"Cache resolution error"

Root Cause: Missing or incorrect cache-dependency-path

Why act Doesn't Catch: act skips caching entirely

Fix:

• uses: actions/setup-node@v4 with: cache: 'npm' cache-dependency-path: package-lock.json # Explicit!

"npm ci requires package-lock.json"

Root Cause: Running npm ci from workspace directory

Why Local Works: May have workspace-specific package-lock.json

Fix:

Run from root

• run: npm ci

Then use working-directory for commands

• working-directory: packages/infra run: pulumi preview

"Service container not starting"

Root Cause: Custom command in options is ignored

Why Local Works: docker-compose uses different container runtime

Fix:

steps:

• run: docker exec [container] sh -c "command &"

Why act Alone Isn't Enough

Local testing with act doesn't catch:

• Cache Validation: act skips GitHub's cache system entirely

• Service Commands: Different container runtime, different behavior

• Build Artifacts: Local has pre-built workspace dependencies

• Clean Environment: CI starts completely fresh every time

• GitHub-Specific Features: Some actions only work on GitHub's runners

Solution: Use the complete validation suite (actionlint + custom validators + act dry run)

Interactive Debugging

When you encounter a workflow failure:

• Share the error: Paste the exact GitHub Actions error message

• I'll analyze: Identify the root cause and why local testing didn't catch it

• I'll fix: Provide corrected workflow configuration

• I'll validate: Show how to test the fix locally

• I'll explain: Clarify why the issue occurred and how to prevent it

Example Interactions

User: "My workflow is failing with 'Some specified paths were not resolved, unable to cache dependencies'"

Me: This is a cache resolution error. You're using cache: 'npm' without specifying cache-dependency-path . GitHub Actions is trying to find package-lock.json but can't locate it.

Here's the fix:

• uses: actions/setup-node@v4 with: node-version: '20' cache: 'npm' cache-dependency-path: package-lock.json # Add this

This wasn't caught by act because it skips caching entirely. To prevent this in the future, run:

.github/scripts/validate-workflows.sh

This script checks for missing cache-dependency-path configurations.

User: "TypeScript can't find '@prpm/types' in CI but works locally"

Me: This is a monorepo build order issue. Your local environment has previously built @prpm/types , but CI starts clean.

Add this before your type check:

• name: Build @prpm/types run: npm run build --workspace=@prpm/types

Why local worked: You've previously run npm run build which compiled @prpm/types to dist/. That compiled output exists in your node_modules. CI doesn't have this.

To test this scenario locally, delete your dist/ folders and try again:

rm -rf packages/*/dist npx tsc --noEmit # Should fail npm run build --workspace=@prpm/types npx tsc --noEmit # Should pass

Continuous Improvement

After each workflow failure in CI:

• Analyze: Why didn't local testing catch this?

• Document: Add to the common failure patterns

• Validate: Update validation scripts to catch it next time

• Test: Ensure the validator actually catches the issue

Best Practices

• Always validate before pushing: Run the complete validation suite

• Keep tools updated: brew upgrade act actionlint yamllint

• Test in clean environment occasionally: Use Docker to simulate fresh CI

• Document failures: Add new patterns to validation scripts

• Use explicit configurations: Never rely on defaults for cache, paths, or commands

Summary

This skill helps you:

• ✅ Catch 90%+ of workflow failures before pushing

• ✅ Understand why local testing didn't catch issues

• ✅ Fix common GitHub Actions problems quickly

• ✅ Build confidence in your CI/CD pipeline

• ✅ Reduce iteration time (no more push-fail-fix-push cycles)

Invoke me whenever you're working with GitHub Actions to ensure your workflows are solid before they hit CI.