skill-fleet 0.1.1

Hierarchical skills taxonomy + DSPy workflow prototype

Skills Fleet

A modular AI capability platform that keeps agent skills organized, discoverable, and standards-compliant.

Skills Fleet lets you create, manage, and deploy AI agent skills as modular, reusable components. Instead of bloated monolithic prompts, skills are organized in a hierarchical taxonomy that agents can load on-demand.

Perfect for: AI development teams building agent systems, platform engineers managing AI capability libraries, and organizations standardizing AI knowledge management.

Why Skills Fleet?

For Technical Teams

• DSPy-Powered Optimization: Built on DSPy (a framework for optimizing LLM workflows) with MIPROv2 and GEPA optimizers for reliable, consistent skill generation.
• agentskills.io Compliant: Standard YAML frontmatter ensures skills work across different agent frameworks.
• Production-Ready: FastAPI v2 server with async background jobs and comprehensive testing.

For Decision Makers

• Modular & Maintainable: Skills are versioned, tracked, and independently testable.
• Standards-Based: Open specification compliance prevents vendor lock-in.
• Scalable: Hierarchical taxonomy for organized growth, supporting hundreds of skills.

For Everyone

• Easy to Use: Simple chat interface for creating skills without coding.
• Validated: Automated compliance checking ensures quality.
• Observable: Built-in analytics and usage tracking.

Prerequisites

• Python: 3.12+
• Package Manager: uv
• API Keys: GOOGLE_API_KEY (Gemini 3 Flash is the default model)

Installation

# Clone the repository
git clone https://github.com/Qredence/skill-fleet.git
cd skill-fleet

# Install dependencies
uv sync

# Setup environment variables
cp .env.example .env
# Edit .env and add your GOOGLE_API_KEY

Quick Start

Create your first skill in under 2 minutes:

1. Start the API server:

   uv run skill-fleet serve
   

2. Create a skill via chat (in a new terminal):

   uv run skill-fleet chat "Create a Python decorators skill"
   

3. Review and Promote: The skill is created as a draft. After reviewing it in skills/_drafts/<job_id>, promote it:

   uv run skill-fleet promote <job_id>
   

Core Commands

Command	Description
uv run skill-fleet serve	Start the FastAPI v2 server (required for most operations)
uv run skill-fleet chat	Interactive conversational skill creation
uv run skill-fleet list	List all skills in the taxonomy
uv run skill-fleet promote <id>	Promote a draft skill to the permanent taxonomy
uv run skill-fleet validate <path>	Validate a skill against agentskills.io standards
uv run skill-fleet generate-xml	Generate an XML registry for agent discovery
uv run skill-fleet optimize	Run DSPy MIPROv2/GEPA optimization
uv run skill-fleet analytics	View usage and performance statistics

Project Structure

skill-fleet/
├── src/skill_fleet/
│   ├── agent/          # Conversational agent for skill creation
│   ├── api/            # FastAPI v2 REST API & routes
│   ├── cli/            # Typer-based CLI (fleet-agent)
│   ├── core/           # Core logic (DSPy programs, tools, models)
│   ├── llm/            # LLM configuration and DSPy setup
│   ├── taxonomy/       # Skill taxonomy management
│   ├── validators/     # Skill validation logic
│   └── workflow/       # DSPy-powered skill creation workflow
├── skills/             # Hierarchical skills taxonomy storage
├── tests/              # Unit and integration tests
├── config/             # Configuration files (config.yaml, templates)
├── scripts/            # Utility and maintenance scripts
└── docs/               # Comprehensive documentation

Configuration

The system is configured via config/config.yaml. This file defines:

• Models: Default model is gemini/gemini-3-flash-preview.
• Roles: Router, Planner, Worker, Judge.
• Optimizers: Settings for MIPROv2 and GEPA.

Environment Variables

• GOOGLE_API_KEY: Required for Gemini models.
• SKILL_FLEET_ENV: (Optional) production (default) or development.
• SKILL_FLEET_CORS_ORIGINS: Required in production. Comma-separated list of allowed origins. Set to * only in development.
• SKILL_FLEET_API_URL: (Optional) Defaults to http://localhost:8000.
• DSPY_TEMPERATURE: (Optional) Override default LLM temperature.

Development & Testing

Running Tests

# Run all tests
uv run pytest

# Run with coverage
uv run pytest --cov=skill_fleet

Linting & Formatting

Uses Ruff for high-performance linting:

uv run ruff check .
uv run ruff format .

Utility Scripts

• scripts/setup_branch_protection.sh: Configure GitHub branch protection.
• scripts/run_dspy_tools.py: Run DSPy optimization and evaluation tools.
• scripts/check_docstrings.py: Verify documentation completeness.

Documentation

• Getting Started
• DSPy Architecture
• CLI Reference
• API Endpoints

License

Proprietary - Copyright (c) 2026 Qredence.

Contributing

Please see CONTRIBUTING.md for details on our development workflow.