Skills vs Subagents vs Hooks: When to Use Each in Claude Code
Skills vs subagents vs hooks in Claude Code, answered: what each one does, context cost, when to pick which, how they combine, plus CLAUDE.md, MCP and plugins.
12 resources tagged “hooks”.
Skills vs subagents vs hooks in Claude Code, answered: what each one does, context cost, when to pick which, how they combine, plus CLAUDE.md, MCP and plugins.
A Claude kit is your bundle of skills, subagents, hooks, MCP servers and prompts. What goes in one, what happened to Claudekit.cc, and how to build your own.
CC Safety Net is a Claude Code plugin hook that blocks destructive commands like git reset --hard and rm -rf, plus access to secrets, before a tool call runs.
Claude Code Templates (aitmpl.com) is an npx CLI and catalog to install agents, commands, hooks, settings, MCPs and skills, plus analytics. Uses, caveats.
Claude-Mem captures what Claude does in each session, compresses it into searchable observations and injects relevant context into later sessions.
Context Mode is a Claude Code plugin that runs tool output through sandboxed tools and a local search index so raw data stays out of your context window.
ECC, formerly everything-claude-code, is an MIT plugin with 71 agents, 311 skills, hooks, rules and AgentShield scanning for Claude Code. Install and caveats.
Anthropic's explanatory-output-style plugin restores the old Explanatory style via a SessionStart hook, so Claude adds short Insight notes as it codes.
Hookify is Anthropic's Claude Code plugin that turns plain-language requests into regex-based hook rules stored as Markdown files that warn on or block actions.
Plannotator opens Claude Code plans, docs and diffs in a local browser UI so you can annotate them and send feedback straight back to the agent. Install, pros.
plugin-dev is Anthropic's toolkit for building Claude Code plugins: seven skills on hooks, MCP, commands and agents, plus a guided create-plugin workflow.
ralph-loop is Anthropic's plugin for the Ralph Wiggum technique: a Stop hook re-feeds one prompt so Claude keeps iterating until a completion promise appears.