Skip to main content

Overview

The plugin exposes 4 tools that the LLM can invoke automatically when you ask natural language questions. You don’t need to call these tools directly — just ask the AI what you want to know.
All tools are read-only. They never modify your filesystem.

search_agents

Search the agent registry by name, description, category, or tag. Returns matching agents with their category and description.

When to Use

Use this tool when looking for agents that match a specific skill, technology, or keyword. Natural language triggers:
  • “Find agents for TypeScript”
  • “Search for Docker specialists”
  • “What agents handle security auditing?”

Parameters

query
string
required
Search query — matches against agent name, description, category, and tags.Examples: "typescript", "docker", "security"

Returns

A formatted string with matching agents:
Found 3 agent(s) matching "react":

• react-specialist (web) — Expert React developer focused on...
  tags: frontend, ui, hooks

• nextjs-developer (web) — Next.js specialist for...
  tags: react, ssr, routing

• ui-designer (web) — UI/UX design specialist...
  tags: design, react, figma
If no agents match, suggests trying a broader term or checking available categories.

Usage Examples

You: Find agents for Python

AI: [invokes search_agents(query="python")]
Found 1 agent(s) matching "python":

• python-pro (languages) — Expert Python developer focused on...
  tags: backend, scripting, data

list_agents

List agents in the registry. Without filters, shows all agents grouped by category with counts. Use category or pack filter for detailed view.
Cannot use both category and pack filters at once.

When to Use

Use this tool when you want to see all available agents, or filter by category/pack. Natural language triggers:
  • “List all agents”
  • “Show agents in the AI category”
  • “What’s in the backend pack?”

Parameters

category
string
Filter by category ID.Valid values: "web", "ai", "devops", "languages", "data-api", "devtools", "security", "mcp", "business", "docs"
pack
string
Filter by pack name.Examples: "frontend", "backend", "startup", "devops"

Returns

Without filters — compact overview:
📦 Agent Registry — 69 agents across 10 categories

💻 Languages (11) — Programming language specialists
  typescript-pro, python-pro, golang-pro, rust-pro, ...

🤖 AI (9) — AI engineering and data science
  ai-engineer, data-scientist, ml-engineer, llm-architect, ...

...
With category filter — detailed list:
🔒 Security — 5 agents
Security auditing and compliance

• security-auditor — Comprehensive security audit specialist
• penetration-tester — Penetration testing and vulnerability assessment
• smart-contract-auditor — Blockchain smart contract security
• security-engineer — Application security engineering
• compliance-auditor — Regulatory compliance and standards
With pack filter — detailed list:
📦 Pack: Backend — 8 agents
Complete backend development stack

• postgres-pro — PostgreSQL database expert
• redis-specialist — Redis caching and data structures
• database-architect — Database design and optimization
• api-architect — RESTful and GraphQL API design
• python-pro — Expert Python developer
• typescript-pro — TypeScript and Node.js specialist
• debugger — Debugging and troubleshooting
• test-automator — Test automation and CI/CD

Usage Examples

You: List all available agents

AI: [invokes list_agents()]
📦 Agent Registry — 69 agents across 10 categories

💻 Languages (11) — Programming language specialists
  typescript-pro, python-pro, golang-pro, ...

🤖 AI (9) — AI engineering and data science
  ai-engineer, data-scientist, ml-engineer, ...

...

get_agent

Get complete details about a specific agent: description, tags, category, installation mode, and current install status.
If you provide a typo or partial name, this tool suggests alternatives.

When to Use

Use this tool when you want full details on a specific agent. Natural language triggers:
  • “Tell me about typescript-pro”
  • “What does the debugger agent do?”
  • “Details on postgres-pro”

Parameters

name
string
required
Exact agent name (kebab-case).Examples: "typescript-pro", "debugger", "react-specialist"

Returns

A formatted string with complete agent details:
typescript-pro

Category:    languages (💻 Languages)
Mode:        primary
Status:      ✓ installed
Description: Expert TypeScript developer focused on type-safe code...
Tags:        typescript, node, backend

Included in packs: backend, frontend, fullstack, startup
Status values:
  • ✓ installed — agent is installed and matches the registry hash
  • ⚠ outdated — agent is installed but content has been modified
  • ✗ not installed — agent is in the registry but not on disk
  • ? unknown — agent is on disk but not tracked in the lock file

Usage Examples

You: What does typescript-pro do?

AI: [invokes get_agent(name="typescript-pro")]
typescript-pro

Category:    languages (💻 Languages)
Mode:        primary
Status:      ✓ installed
Description: Expert TypeScript developer focused on type-safe code,
             modern ECMAScript features, and Node.js best practices.
Tags:        typescript, node, backend

Included in packs: backend, frontend, fullstack, startup

check_health

Health check on agent installation. Reports installed, outdated, missing, and unknown agents. Verifies file integrity against lock file.
This tool may take a moment to complete if you have many agents installed, as it computes SHA256 hashes for each file.

When to Use

Use this tool when you want to verify your agent installation status. Natural language triggers:
  • “Check agent health”
  • “Are my agents up to date?”
  • “Verify agent installation”

Parameters

None. This tool takes no arguments.

Returns

A formatted health report:
Agent Health Report

Summary:
  ✓ 23 installed (hashes match)
  ⚠ 2 outdated (content modified since install)
  ✗ 46 not installed
  ? 1 unknown (on disk but not tracked)

Outdated agents:
  • typescript-pro
  • debugger

Unknown agents:
  • custom-agent

Integrity: ✓ 23 OK · ⚠ 2 mismatched · ✗ 0 missing from disk
Summary fields:
  • Installed — agents match the registry hash (no modifications)
  • Outdated — agents on disk have been modified since installation
  • Not installed — agents in the registry but not on disk
  • Unknown — agents on disk but not tracked in the lock file (custom agents)
Integrity:
  • OK — lock entry exists and hash matches
  • Mismatched — lock entry exists but hash doesn’t match (file was edited)
  • Missing from disk — lock entry exists but file is gone

Usage Examples

You: Check if my agents are up to date

AI: [invokes check_health()]
Agent Health Report

Summary:
  ✓ 23 installed (hashes match)
  ⚠ 2 outdated (content modified since install)
  ✗ 46 not installed
  ? 0 unknown (on disk but not tracked)

Outdated agents:
  • typescript-pro
  • debugger

Integrity: ✓ 23 OK · ⚠ 2 mismatched · ✗ 0 missing from disk

You may want to reinstall the outdated agents to get the latest versions.

Response Fields

All tools return formatted strings, not JSON. This is intentional — the LLM reads the formatted output and presents it to you naturally.
Formatted String
string
Human-readable text with ANSI-like formatting (icons, indentation, bullet points).The LLM interprets this output and rephrases it conversationally.

Error Handling

All tools sanitize errors to avoid leaking filesystem structure:
  • Absolute paths are replaced with …/ or …\
  • Windows drives (C:\) become …\
  • UNC paths (\\server\) become …\
  • Deep Unix paths (/a/b/c/d/e) become …/
You’ll never see your full filesystem path in tool output.

Next Steps

Installation

Install the plugin globally or per-project

Plugin Overview

Learn how the plugin works

Build docs developers (and LLMs) love