Skills System

The Bindu Skills System provides rich agent capability advertisement for intelligent orchestration and agent discovery. Inspired by Claude’s skills architecture, it enables agents to provide detailed documentation about their capabilities for orchestrators to make informed routing decisions.

What are Skills?

Skills in Bindu serve as rich advertisement metadata that help orchestrators:

🔍 Discover the right agent for a task
📖 Understand detailed capabilities and limitations
✅ Validate requirements before execution
📊 Estimate performance and resource needs
🔗 Chain multiple agents intelligently

Skills are not executable code—they’re structured metadata that describe what your agent can do.

Architecture

How It Works

Agent defines skills - YAML files describe capabilities
Skills API exposes - REST endpoints provide skill information
Orchestrator discovers - Queries available skills
Capability matching - Finds agents that meet requirements
Task routing - Sends tasks to optimal agent

Why Skills Matter

For Agent Developers

Clear documentation - Define capabilities in one place
Standardized format - YAML-based, human-readable
Easy maintenance - Update skills without code changes
Better discovery - Help users find your agent

For Orchestrators

Intelligent routing - Match tasks to the right agent
Capability validation - Check requirements before execution
Performance estimation - Predict resource needs
Multi-agent workflows - Chain agents based on capabilities

For Users

Transparency - Understand what agents can do
Reliability - Know limitations upfront
Better results - Tasks routed to optimal agents

Skill Definition Format

Bindu uses YAML files as the single source of truth for skill definitions.

Benefits of YAML

✅ Human-readable - Easy to write and understand
✅ Structured data - Programmatic access for orchestrators
✅ Rich documentation - Everything in one place
✅ LLM-friendly - Easy for AI to parse and understand
✅ No duplication - Single source of truth

Directory Structure

examples/skills/
├── text-analysis/
│   └── skill.yaml
├── image-generation/
│   └── skill.yaml
└── data-processing/
    └── skill.yaml

Basic Skill Example

Here’s a minimal skill definition:

# Basic Metadata
id: text-analysis-v1
name: text-analysis
version: 1.0.0
author: [email protected]

# Description
description: |
  Analyzes text for sentiment, entities, and key phrases.
  Supports multiple languages and formats.

# Tags
tags:
  - nlp
  - text-analysis
  - sentiment

# Input/Output Modes
input_modes:
  - text/plain
  - application/json

output_modes:
  - application/json

# Example Queries
examples:
  - "Analyze the sentiment of this review"
  - "Extract entities from this document"
  - "Identify key phrases in this text"

# Capabilities
capabilities_detail:
  sentiment_analysis:
    supported: true
    types:
      - positive
      - negative
      - neutral
    limitations: "Supports English only"

  entity_recognition:
    supported: true
    types:
      - person
      - organization
      - location

# Requirements
requirements:
  packages:
    - transformers>=4.0.0
    - torch>=2.0.0
  min_memory_mb: 512

# Performance
performance:
  avg_processing_time_ms: 2000
  max_file_size_mb: 10
  concurrent_requests: 5

Complete Skill Structure

A full skill definition includes:

1. Basic Metadata

id: skill-id-v1
name: skill-name
version: 1.0.0
author: [email protected]
description: |
  What this skill does

2. Tags and Modes

tags:
  - category
  - feature

input_modes:
  - text/plain
  - application/json

output_modes:
  - application/json

3. Capabilities Detail

capabilities_detail:
  feature_1:
    supported: true
    types:
      - type_a
      - type_b
    limitations: "Optional limitations"

  feature_2:
    supported: false
    planned_version: "2.0.0"

4. Requirements

requirements:
  packages:
    - package1>=1.0.0
    - package2>=2.0.0
  system:
    - system-dependency
  min_memory_mb: 512

5. Performance Metrics

performance:
  avg_processing_time_ms: 2000
  max_file_size_mb: 50
  concurrent_requests: 5
  memory_per_request_mb: 500

6. Documentation

documentation:
  overview: |
    Comprehensive description of the skill

  use_cases:
    when_to_use:
      - Scenario 1
      - Scenario 2
    when_not_to_use:
      - Anti-pattern 1

  input_structure: |
    Expected input format with examples

  output_format: |
    Output format with examples

  error_handling:
    - "Error type 1: How it's handled"

  examples:
    - title: "Example 1"
      input: "input data"
      output: "output data"

  best_practices:
    for_developers:
      - "Best practice 1"
    for_orchestrators:
      - "Routing guideline 1"

Creating a New Skill

Create Directory

mkdir -p examples/skills/my-skill

Create skill.yaml

cd examples/skills/my-skill
touch skill.yaml

Define Skill

id: my-skill-v1
name: my-skill
version: 1.0.0
author: [email protected]

description: |
  What your skill does

tags:
  - relevant
  - tags

input_modes:
  - text/plain

output_modes:
  - application/json

capabilities_detail:
  main_feature:
    supported: true
    description: "What it does"

requirements:
  packages: []
  min_memory_mb: 100

performance:
  avg_processing_time_ms: 500
  concurrent_requests: 10

documentation:
  overview: |
    Detailed description

  use_cases:
    when_to_use:
      - Use case 1
    when_not_to_use:
      - Anti-pattern 1

Reference in Config

Add to your agent config:

{
  "skills": [
    "examples/skills/my-skill"
  ]
}

Test

# Start agent
python your_agent.py

# Test endpoints
curl http://localhost:3773/agent/skills
curl http://localhost:3773/agent/skills/my-skill-v1
curl http://localhost:3773/agent/skills/my-skill-v1/documentation

API Endpoints

List All Skills

GET /agent/skills

Response:

{
  "skills": [
    {
      "id": "text-analysis-v1",
      "name": "text-analysis",
      "version": "1.0.0",
      "description": "Analyzes text for sentiment...",
      "tags": ["nlp", "text-analysis"]
    }
  ]
}

Get Skill Details

GET /agent/skills/{skill_id}

Response:

{
  "id": "text-analysis-v1",
  "name": "text-analysis",
  "version": "1.0.0",
  "capabilities_detail": {...},
  "requirements": {...},
  "performance": {...}
}

Get Skill Documentation

GET /agent/skills/{skill_id}/documentation

Response:

{
  "overview": "Comprehensive description...",
  "use_cases": {...},
  "examples": [...]
}

Orchestrator Integration

Discovery Example

import requests

# Discover available skills
response = requests.get("http://agent:3773/agent/skills")
skills = response.json()["skills"]

# Find skills with specific tags
nlp_skills = [s for s in skills if "nlp" in s["tags"]]

# Get detailed capabilities
for skill in nlp_skills:
    details = requests.get(
        f"http://agent:3773/agent/skills/{skill['id']}"
    ).json()
    
    if details["capabilities_detail"]["sentiment_analysis"]["supported"]:
        print(f"Found sentiment analysis: {skill['name']}")

Capability Matching

def find_agent_for_task(task_requirements):
    """Find best agent based on requirements."""
    agents = discover_agents()
    
    for agent in agents:
        skills = get_agent_skills(agent)
        
        for skill in skills:
            if matches_requirements(skill, task_requirements):
                return agent, skill
    
    return None, None

LLM-Based Orchestration

# Provide skills to LLM for routing
skills_context = """
Available agents and their skills:

1. Text Analysis Agent
   - Sentiment analysis (English only)
   - Entity recognition
   - Key phrase extraction
   - Avg time: 2s, Max size: 10MB

2. Translation Agent
   - 50+ language pairs
   - Document translation
   - Avg time: 5s, Max size: 5MB
"""

# LLM can now make informed routing decisions

Best Practices

For Skill Authors

✅ Be specific - Clearly define capabilities and limitations
✅ Include examples - Show real usage scenarios
✅ Document errors - Explain how errors are handled
✅ Update versions - Increment version on changes
✅ Test thoroughly - Validate YAML and endpoints

For Agent Developers

✅ Keep skills updated - Reflect actual capabilities
✅ Use semantic versioning - Major.Minor.Patch
✅ Provide rich documentation - Help orchestrators understand
✅ Include performance metrics - Enable capacity planning
✅ Tag appropriately - Improve discoverability

For Orchestrators

✅ Cache skill data - Reduce API calls
✅ Validate before routing - Check requirements match
✅ Consider performance - Route based on metrics
✅ Handle failures - Have fallback strategies
✅ Monitor usage - Track which skills are used

Configuration

File-Based Skills (Recommended)

{
  "name": "my-agent",
  "skills": [
    "examples/skills/text-analysis",
    "examples/skills/data-processing"
  ]
}

Inline Skills (Legacy)

{
  "name": "my-agent",
  "capabilities": {
    "text-analysis": {
      "description": "Analyzes text",
      "tags": ["nlp"]
    }
  }
}

Mixed Approach

{
  "name": "my-agent",
  "skills": [
    "examples/skills/text-analysis"
  ],
  "capabilities": {
    "legacy-skill": {
      "description": "Old format"
    }
  }
}

Troubleshooting

Skill Not Loading

Check:

Path is correct in config
skill.yaml exists in directory
YAML syntax is valid
Agent has read permissions

Invalid YAML

Solution:

Validate YAML syntax online
Check indentation (use spaces, not tabs)
Ensure all required fields are present
Use quotes for special characters

Missing Documentation

Cause: Documentation section not definedSolution: Add documentation section to skill.yaml

Fields Not Showing

Cause: Fields not exposed in APISolution: Check API endpoint returns expected fields

Examples

Text Analysis Skill

id: text-analysis-v1
name: text-analysis
version: 1.0.0
author: [email protected]

description: |
  Advanced text analysis with sentiment, entities, and key phrases

tags:
  - nlp
  - sentiment
  - entities

capabilities_detail:
  sentiment_analysis:
    supported: true
    types: [positive, negative, neutral]
    languages: [en]
  
  entity_recognition:
    supported: true
    types: [person, org, location]

Image Generation Skill

id: image-gen-v1
name: image-generation
version: 1.0.0
author: [email protected]

description: |
  Generate images from text descriptions

tags:
  - image
  - generation
  - ai-art

capabilities_detail:
  text_to_image:
    supported: true
    styles: [realistic, artistic, cartoon]
    max_resolution: "1024x1024"

performance:
  avg_processing_time_ms: 15000
  max_concurrent_requests: 2

Next Steps

Create Your First Skill

Step-by-step guide to creating skills

Skill Reference

Complete YAML field reference

Orchestration Guide

Build multi-agent workflows

GitHub Examples

View example skills

Resources

Skills System Docs - Complete technical documentation
Example Skills - Real-world skill definitions
YAML Validator - Validate your YAML syntax
Discord Community - Get help and share skills

Introduction

Create Bindu Agent in 2 minutes

Concepts

Skills

Learn

How to

​Skills System

​What are Skills?

​Architecture

​How It Works

​Why Skills Matter

​For Agent Developers

​For Orchestrators

​For Users

​Skill Definition Format

​Benefits of YAML

​Directory Structure

​Basic Skill Example

​Complete Skill Structure

​1. Basic Metadata

​2. Tags and Modes

​3. Capabilities Detail

​4. Requirements

​5. Performance Metrics

​6. Documentation

​Creating a New Skill

​API Endpoints

​List All Skills

​Get Skill Details

​Get Skill Documentation

​Orchestrator Integration

​Discovery Example

​Capability Matching

​LLM-Based Orchestration

​Best Practices

​For Skill Authors

​For Agent Developers

​For Orchestrators

​Configuration

​File-Based Skills (Recommended)

​Inline Skills (Legacy)

​Mixed Approach

​Troubleshooting

​Examples

​Text Analysis Skill

​Image Generation Skill

​Next Steps

Create Your First Skill

Skill Reference

Orchestration Guide

GitHub Examples

​Resources

Skills System

What are Skills?

Architecture

How It Works

Why Skills Matter

For Agent Developers

For Orchestrators

For Users

Skill Definition Format

Benefits of YAML

Directory Structure

Basic Skill Example

Complete Skill Structure

1. Basic Metadata

2. Tags and Modes

3. Capabilities Detail

4. Requirements

5. Performance Metrics

6. Documentation

Creating a New Skill

API Endpoints

List All Skills

Get Skill Details

Get Skill Documentation

Orchestrator Integration

Discovery Example

Capability Matching

LLM-Based Orchestration

Best Practices

For Skill Authors

For Agent Developers

For Orchestrators

Configuration

File-Based Skills (Recommended)

Inline Skills (Legacy)

Mixed Approach

Troubleshooting

Examples

Text Analysis Skill

Image Generation Skill

Next Steps

Resources