More than 20 breaches in 31 days. The largest crypto heist in history. A ransomware wave that reached NAS devices in Vietnam. And one common thread that is more worrying than any single number.

Finished the PCS AI Learning Roadmap and want to go deeper? This post rounds up free (or nearly free) AI courses curated by PCS Vietnam from Google, IBM, Microsoft, and DeepLearning.AI — grouped by role so you can find the right next step for you.

Before you dive in: These are third-party courses — each has its own approach and terminology. PCS recommends completing the AI Roadmap first so you have a solid conceptual foundation and can pick the right path instead of getting overwhelmed by hundreds of options.

Based on exploring the comprehensive Everything Claude Code repository — 156 skills, 72 commands, 38 agents. This guide covers how to cherry-pick tools without installing the full package.

Repository Overview​

The Everything Claude Code repository is an Anthropic Hackathon winner containing 156 reusable skills across multiple AI platforms.

Scope: 156 skills | 72 commands | 38 agents | 12+ hooks

Target platforms: Claude Code (primary), Cursor, Codex, Gemini

Claude Code-Only vs Shared Access​

Claude Code-Only Components​

These features exist in ~/.claude/ and are not accessible to other AI environments:

• ~/.claude/skills/ — workflow definitions and automation scripts
• ~/.claude/commands/ — slash command shims and integrations
• ~/.claude/agents/ — subagent configurations
• settings.json hooks — system event handlers

Shared Components (Multi-AI)​

These can be accessed by Antigravity Agent, Gemini, or other tools:

• rules/common/*.md — security and style guidelines
• mcp-configs/ — JSON Model Context Protocol server configurations

Critical Hook Dependency Issue​

A major finding: ~90% of hooks depend on ${CLAUDE_PLUGIN_ROOT}/scripts/hooks/*.js

This means:

• You cannot cherry-pick individual hooks
• To use the full hooks system, you must run install.ps1 --profile full (complete installation)
• Only simple hooks like block-no-verify work standalone

Implication: If you want selective skill cherry-picking without the full installation overhead, focus on skills and commands first.

Recommended Cherry-Pick Strategy​

For a tech stack focused on JavaScript / React / Cloudflare / PowerShell, here's what to cherry-pick:

Priority 1: Skills → ~/.claude/skills/​

Priority 2: Commands → ~/.claude/commands/​

Priority 3: Agents → ~/.claude/agents/​

Priority 4: Rules → Merge into CLAUDE.md​

• rules/common/security.md — standard security checklist for all code
• rules/common/coding-style.md — immutability and file organization patterns

Installation Command (Selective)​

# Copy skills
cp -r Research/Claudeskill/skills/security-review ~/.claude/skills/
cp -r Research/Claudeskill/skills/mcp-server-patterns ~/.claude/skills/
cp -r Research/Claudeskill/skills/api-design ~/.claude/skills/
cp -r Research/Claudeskill/skills/frontend-patterns ~/.claude/skills/
cp -r Research/Claudeskill/skills/deployment-patterns ~/.claude/skills/
cp -r Research/Claudeskill/skills/database-migrations ~/.claude/skills/
cp -r Research/Claudeskill/skills/git-workflow ~/.claude/skills/

# Copy commands
cp Research/Claudeskill/commands/code-review.md ~/.claude/commands/
cp Research/Claudeskill/commands/security-scan.md ~/.claude/commands/
cp Research/Claudeskill/commands/build-fix.md ~/.claude/commands/
cp Research/Claudeskill/commands/plan.md ~/.claude/commands/

# Copy agents
cp -r Research/Claudeskill/agents/code-reviewer ~/.claude/agents/
cp -r Research/Claudeskill/agents/security-reviewer ~/.claude/agents/

# Copy and merge rules
cp Research/Claudeskill/rules/common/security.md ~/.claude/docs/SECURITY.md
cp Research/Claudeskill/rules/common/coding-style.md ~/.claude/docs/STYLE.md

What to Skip​

Language-specific skills (if not in your stack):

• All Python, Java, Kotlin, Swift, Android skills
• Rust, PHP, Perl modules
• Language-specific build and test commands

Complex orchestration (too heavy for solo operators):

• Full hooks system (requires Node.js runtime + complete installation)
• Agent harness orchestration (intended for team workflows)

Key Learnings​

1. Hooks require full installation — cherry-pick skills instead
2. Security and API design should be top priority for pcs-platform work
3. Docusaurus patterns are essential for jaxvn-blog maintenance
4. Git workflow rules should be merged into project CLAUDE.md

Related Notes​

• [[guide-2026-04-04-cloudflare-auth-worker]] — pcs-platform Phase 2a implementation
• [[2026-04-03]] — daily notes from initial pcs-auth Worker build
• [[note-2026-04-04-everything-claude-code]] — vault original (Vietnamese)

Status: ✅ Published 2026-04-12 | Integration planned for Phase 2g

Most Vietnamese SMBs don't have Intune or SCCM. They have a mix of domain-joined PCs, workgroup machines, and sometimes both on the same floor. But you can still enforce consistent browser policies across all of them — using nothing but registry keys and a PowerShell script.

Everything Claude Code is an Anthropic Hackathon-winning repo with 50K+ stars. It ships 156 skills, 72 commands, 38 agents, and a full hooks system for Claude Code. Here's my honest breakdown after auditing it for my own stack.

Congratulations, you have made your first post!

Feel free to play around and edit this post as much as you like.