extend
Add one new capability to an existing ALBA setup - skill, hook, rule, slash command or MCP integration - interactively, then wire it into CLAUDE.md and settings.json. Use when user says "add a skill", "add a hook", "I want a rule for X", "connect Trello/Gmail/Linear", "extend my agent", "can ALBA also do X". Do NOT use for first-time setup (use /setup). If the user already knows they want a skill, /create-skill goes straight to writing SKILL.md - /extend is the router for when the component type is still open.
onurpolat05 · Static checks on file
Installation
Read the source and dependencies before running a command. Copying does not install or execute anything.
npx skills add onurpolat05/ALBA --skill extendAbout this skill
name: extend description: Add one new capability to an existing ALBA setup - skill, hook, rule, slash command or MCP integration - interactively, then wire it into CLAUDE.md and settings.json. Use when user says "add a skill", "add a hook", "I want a rule for X", "connect Trello/Gmail/Linear", "extend my agent", "can ALBA also do X". Do NOT use for first-time setup (use /setup). If the user already knows they want a skill, /create-skill goes straight to writing SKILL.md - /extend is the router for when the component type is still open. context: inline effort: medium argument-hint: "[skill|hook|rule|command|mcp]" allowed-tools: [Read, Write, Edit, Glob, Grep, AskUserQuestion]
ALBA Extend
Add new capabilities to the agent system. Interactive and collaborative.
Step 1: Identify
Ask: "What would you like to add to your agent?"
Types:
- Skill - New capability (research, content creation, task automation...)
- Hook - Automated event response (session start, error logging...)
- Rule - Behavioral guideline (code standards, communication style...)
- Command - Custom /slash command
- MCP Integration - Connect external tool (Trello, Gmail, Calendar...)
Step 2: Build
Skills
- Understand: What task? What inputs/outputs? What tools?
- Check
templates/skills/for existing templates - Create
.claude/skills/[name]/SKILL.mdwith frontmatter:
The--- name: [skill-name] description: [what it does]. Use when user says "x", "y". Do NOT use for [z]. context: fork # fork = subagent (protects context), inline = main context background: false # fork only: wait for the result in this turn (fork defaults to background) allowed-tools: [tools] ---descriptionis the only field that decides whether the skill fires at the right moment — a bare one-line summary will not trigger reliably. Full field reference:templates/skills/SKILL-TEMPLATE.md. - Add to CLAUDE.md Skills table
- Test immediately
Hooks
- Identify the event. Common ones:
SessionStart,UserPromptSubmit,PreToolUse,PostToolUse,PostToolUseFailure,Stop,SubagentStop,PreCompact,PostCompact,SessionEnd. Claude Code v2.1.220 defines 30 events — seetemplates/hooks/README-hooks.mdfor the full list and each one's output contract. - Define what should happen when triggered
- Create
.claude/hooks/[name].shand make executable (chmod +x) - Add to
.claude/settings.json— note the nestedhooksarray; skipping the inner level is the most common reason a newly added hook never runs:{ "hooks": { "[EventType]": [ { "matcher": "Bash", "hooks": [ { "type": "command", "command": "${CLAUDE_PROJECT_DIR}/.claude/hooks/[name].sh", "timeout": 5 } ] } ] } }matcheronly applies to tool events. Use${CLAUDE_PROJECT_DIR}rather than a relative path, or the hook breaks for sessions started from a subdirectory. - Test by triggering the event
See templates/hooks/ for examples.
Rules
- Understand: When does this apply? What to do/not do? Why?
- Classify: behavioral, security, process, quality, restriction
- Create
.claude/rules/[name].md - Rules auto-load from
.claude/rules/- no CLAUDE.md update needed - Verify rule is followed
See templates/rules/ for examples.
Commands
- Define: name, purpose, parameters
- Create
.claude/commands/[name].mdwith description frontmatter - Test with /[name]
MCP Integrations
- Check if MCP server is already configured (try listing tools)
- If not configured: guide through authentication setup
- Create helper skill for common operations
- Add to CLAUDE.md Tools & Integrations table
- Test the connection
Step 3: Verify
- Does it work as expected? (test with real example)
- Is CLAUDE.md updated if needed?
- Does user understand how to use it?
Guidelines
- Check existing skills/hooks/rules before creating new ones
- Use
templates/as starting points when available - Keep it simple - solve the specific need, don't over-engineer
- Test each component immediately after creation