From plugin-dev
Use when the user asks to "create a slash command", "add a command", "create a hook", "add a PreToolUse/PostToolUse hook", "create an agent", "write a subagent", "add MCP server", "integrate MCP", "configure MCP in plugin", "use .mcp.json", or needs guidance on any plugin component: commands, hooks, agents, MCP servers, or output styles.
npx claudepluginhub karimstekelenburg/k3m-marketplace --plugin plugin-devThis skill uses the workspace's default tool permissions.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Guides implementation of event-driven hooks in Claude Code plugins using prompt-based validation and bash commands for PreToolUse, Stop, and session events.
Create and configure Claude Code plugin components — commands, hooks, agents, MCP servers, and output styles.
Before answering ANY question about Claude Code component capabilities, syntax, or configuration:
firecrawl map https://code.claude.com/docs/en/ to see all available pages (optionally add --search "<component-type>" to narrow)firecrawl scrape <url> -f markdown -o .firecrawl/<topic>.md to get the full pageKey doc pages to search for: commands, hooks, agents, mcp, output-styles, plugins-reference
commands/*.md)Slash commands are markdown files with optional YAML frontmatter. They are instructions FOR Claude, not programs.
Key frontmatter fields: description, allowed-tools, model, effort, argument-hint, disable-model-invocation
Features:
$ARGUMENTS and positional args ($1, $2) for dynamic input@path/to/file syntax to include file contents!`command` for inline bash executionhooks/hooks.json)Event-driven automation. Plugin format wraps with {"hooks": {EVENT: [...]}}.
Hook types: command, prompt, agent, http
Key events: PreToolUse, PostToolUse, Stop, SessionStart, SessionEnd, PreCompact, PostCompact, UserPromptSubmit, Notification, SubagentStop
Important:
${CLAUDE_PLUGIN_ROOT} in hook commands for portability$CLAUDE_ENV_FILE in SessionStart hooks to persist env varspermissionDecision: allow|deny|askagents/*.md)Subagent definitions with YAML frontmatter and system prompt body.
Required frontmatter: name, description, model, color
Optional: effort, maxTurns, disallowedTools, tools, initialPrompt, hooks
Key patterns:
<example> blocks for triggeringhooks field allows inline hook definitions scoped to agent lifetimeinitialPrompt auto-submits first turn.mcp.json)External tool integration via Model Context Protocol.
Server types: stdio (local process), http (streamable HTTP — recommended for remote)
Key patterns:
${CLAUDE_PLUGIN_ROOT} in server commands for portabilitymcp__plugin_<plugin>_<server>__<tool>/mcp command to verify servers loadedoutput-styles/*.md)Response behavior modification via markdown files + SessionStart hook injection.
Always fetch current docs for the latest output styles specification.
${CLAUDE_PLUGIN_ROOT} for all internal path referencesdocs/file-roles-and-dependencies.md)<example> blocks need description: | block scalar for YAML validation