Fleet Skills — Agent Skills Collection

This repository serves as a centralized collection of agent skills, designed to be compatible with multiple AI agent ecosystems including Claude Code and Cursor.

Quick Start — Install a Skill

Install skills directly from this repository using npx:

# Install all skills from the repository
npx skills add Qredence/skills

# Install globally
npx skills add -g Qredence/skills

Available Skills

Skill	Description
fastapi-router-py	FastAPI router utilities
dspy-core	DSPy core patterns
dspy-development	DSPy development workflows
dspy-fleet-rlm	DSPy fleet RLM integration
dspy-gepa	DSPy Gepa patterns
dspy-optimization	DSPy optimization techniques
dspy-signature	DSPy signature design
agent-converter	Agent format converter
babysit-pr	PR babysitting automation
rlm	fleet-rlm / daytona_pilot runtime model
rlm-batch	Batched and parallel RLM execution
rlm-debug	fleet-rlm runtime diagnostics
rlm-execute	Daytona sandbox execution with persistence
rlm-long-context	Long-context RLM processing (experimental)
rlm-memory	Daytona durable volume memory patterns
rlm-run	fleet-rlm entrypoint selection
rlm-test-suite	fleet-rlm test lane guidance

Repository Structure

skills/                  # Single source of truth for skill definitions
├── fastapi-router-py/  # Individual skill directories
├── dspy-core/
└── ...

plugins/                # Ecosystem-specific packaging
├── fleet-skills/       # Main plugin package
│   ├── .claude-plugin/
│   ├── .cursor-plugin/
│   ├── agents/         # Sub-agent definitions
│   └── skills/         # Symlinked skills

scripts/                # Management utilities
├── init_skill.py               # Scaffold new skills
├── package_skill.py            # Validate & package skills
├── sync_plugins.py             # Sync to plugin directories
└── evaluate_skills_litellm.py  # Real-LLM skill evaluator (LiteLLM proxy)

tests/                  # Skill evaluation harness (pnpm)
├── harness/            # TypeScript evaluator framework
├── scenarios/          # Scenario YAML files per skill (51 RLM scenarios)
└── AGENTS.md           # Testing agent instructions

.github/skills/         # Acceptance criteria per skill

Creating a New Skill

uv run python scripts/init_skill.py <skill-name> --path skills/

This creates:

skills/<skill-name>/
├── SKILL.md
├── scripts/
├── references/
└── assets/

Packaging a Skill

uv run python scripts/package_skill.py skills/<skill-name>

Generates a <skill-name>.skill zip archive if valid.

Contributing

See .github/CONTRIBUTING.md for detailed guidelines on creating and packaging skills.