Tool Discovery Skill

Recommend the right agents and skills for any task. Covers both heavyweight agents (Task tool) and lightweight skills (Skill tool).

Triggers

which agent, which skill, what tool should I use, help me choose, recommend agent, recommend skill, find the right tool, what's available

Decision Flowchart

Is this a reference/lookup task?
├── YES → Use a SKILL (lightweight, auto-injects)
│         Examples: patterns, syntax, CLI commands
│
└── NO → Does it require reasoning/decisions?
         ├── YES → Use an AGENT (heavyweight, spawns subagent)
         │         Examples: architecture, optimization, debugging
         │
         └── MAYBE → Check both lists below

Rule of thumb:

• Skills = Quick reference, patterns, commands (50-200 lines)
• Agents = Deep expertise, autonomous decisions (200-1600 lines)

---

Skills Reference

Pattern Skills (Reference Tables)

Skill	Triggers	Use When
rest-patterns	rest api, http methods, status codes	HTTP method semantics, status code lookup
tailwind-patterns	tailwind, utility classes, tw	Tailwind classes, responsive breakpoints
sql-patterns	sql patterns, cte, window functions	CTE examples, JOIN reference, window functions
sqlite-ops	sqlite, aiosqlite, local database	SQLite schema patterns, Python sqlite3/aiosqlite
mcp-patterns	mcp server, model context protocol	MCP server structure, tool handlers

CLI Tool Skills

Skill	Triggers	Use When
file-search	fd, ripgrep, rg, fzf	Finding files, searching code, interactive selection
find-replace	sd, batch replace	Modern find-and-replace (sd over sed)
code-stats	tokei, difft, line counts	Codebase statistics, semantic diffs
data-processing	jq, yq, json, yaml	JSON/YAML processing and transformation
structural-search	ast-grep, sg, ast pattern	Search by AST structure, not text

Workflow Skills

Skill	Triggers	Use When
git-workflow	lazygit, gh, delta, pr, rebase, stash, bisect	Git operations, GitHub PRs, staging, rebase, bisect
python-env	uv, venv, pip, pyproject	Python environment setup with uv
task-runner	just, justfile, run tests	Running project tasks via justfile
doc-scanner	AGENTS.md, conventions, scan docs	Finding and reading project documentation
project-planner	plan, sync plan, track	Project planning with /plan command

---

Agents Reference

Language Experts

Agent	Use When
python-expert	Advanced Python, async, testing, optimization
javascript-expert	Modern JS, async patterns, V8 optimization
typescript-expert	Type system, generics, complex types
bash-expert	Shell scripting, defensive programming

Framework Experts

Agent	Use When
react-expert	React hooks, Server Components, state management
vue-expert	Vue 3, Composition API, Pinia
laravel-expert	Laravel, Eloquent, PHP testing
astro-expert	Astro SSR/SSG, Cloudflare deployment

Infrastructure Experts

Agent	Use When
postgres-expert	PostgreSQL optimization, execution plans
sql-expert	Complex queries, query optimization
wrangler-expert	Cloudflare Workers deployment
aws-fargate-ecs-expert	ECS/Fargate container orchestration
cloudflare-expert	Workers, Pages, DNS configuration

Specialized

Agent	Use When
firecrawl-expert	Web scraping, crawling, anti-bot bypass
payloadcms-expert	Payload CMS architecture, configuration
craftcms-expert	Craft CMS, Twig templates
cypress-expert	E2E testing, component tests
project-organizer	Restructuring project directories

Built-in Agents (Task tool)

Agent	Use When
Explore	Quick codebase exploration, "where is X"
Plan	Design implementation strategy
general-purpose	Multi-step tasks when unsure
claude-code-guide	Questions about Claude Code features

---

Matching By Context

By File Extension

Files	Skill	Agent
.py	python-env	python-expert
.ts, .js	—	typescript-expert, javascript-expert
.sql	sql-patterns	postgres-expert, sql-expert
.sh	—	bash-expert
.astro	tailwind-patterns	astro-expert
.json	data-processing	—
.yaml	data-processing	—

By Task Type

Task	Try Skill First	Then Agent
"How do I write a CTE?"	sql-patterns	sql-expert
"Optimize this query"	—	postgres-expert
"Find files named X"	file-search	Explore
"Restructure this project"	—	project-organizer
"Scrape this website"	—	firecrawl-expert
"What HTTP status for X?"	rest-patterns	—
"Set up Python project"	python-env	python-expert
"Build MCP server"	mcp-patterns	—

By Keywords

Keywords	Likely Skill	Likely Agent
"pattern", "example", "syntax"	Check skills first	—
"optimize", "debug", "fix"	—	Check agents
"reference", "lookup", "how to"	Check skills first	—
"architecture", "design", "plan"	—	Check agents or Plan

---

How to Launch

Skills (via Skill tool)

Skill tool → skill: "file-search"

Skills auto-inject into current context. No subagent spawned.

Agents (via Task tool)

Task tool → subagent_type: "python-expert"
         → prompt: "Your task description"

Agents spawn a subagent session with their full context.

---

Recommendations Workflow

User: "Which tool should I use for X?"

1. Parse the request:
   - Is it reference/lookup? → Skill
   - Does it need reasoning? → Agent
   - Unclear? → Check both lists

2. Match to available tools:
   - Check file types in project
   - Check config files (package.json, pyproject.toml, etc.)
   - Consider task complexity

3. Output format:
   RECOMMENDED: [skill/agent-name]
   TYPE: Skill | Agent
   WHY: [1 sentence rationale]

   LAUNCH: Skill tool with "name" | Task tool with subagent_type="name"

4. If multiple apply:
   PRIMARY: [name] - [reason]
   SECONDARY: [name] - [reason]

---

Tips

• Skills are cheaper - Use for reference lookups, patterns, CLI commands
• Agents are powerful - Use for decisions, optimization, debugging
• Don't over-recommend - Maximum 2-3 tools per task
• Parallel execution - Launch independent agents in parallel via Task tool
• Check availability - Run /agents or check this skill for current list