An all-in-one guide repository systematically explaining Claude Code's main extensions — Skills, Hooks, MCP, and Subagents (Agent Teams) — has been released, covering everything from setup to advanced workflow automation step by step.
Claude Code · Extension Guide
From Prompt Engineering to System Architecture
A new all-in-one guide maps Claude Code's four core extensions — Skills, Hooks, MCP, and Subagents — plus Plugins, turning Anthropic's terminal-native CLI agent into a designed execution environment rather than a chat box.
5
Extension points: Skills · Hooks · Subagents · MCP · Plugins
6–10h
saved on complex tasks — 12-step spec to code in ~2 hours
1M
token context (beta) on Opus 4.7 · 128K max output
The five extension points
Role · placement / invocation
Skills
Reusable knowledge & workflows
.claude/commands/*.md
Hooks
Event-driven automation scripts
.claude/settings.json
Subagents
Specialized agents in isolated contexts
.claude/agents/*.md
MCP
External DB / API / browser links
per-project config
Plugins
Bundle of all the above, distributable
/plugin install · git/npm
Model lineup & effort levels
Heavier model → more context, more output, higher default effort
Opus 4.7
flagship · 1M ctx
Haiku 4.5
lightweight tasks
Effort range: low → high → max · controls depth of reasoning
An automation loop in practice
Subagent explores UI
→
Playwright MCP drives browser
→
Generate tests
→
Debug & fix
Where it shines
"Strongest coding agent depending on setup"
Terminal-native, editor-independent, tight git/shell integration
Big time savings on multi-step, complex work
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…