ORCA Agent Skills

Frameworks & SDKs 💻 Python ⚖️ Apache-2.0 🟡 Quiet lately
61 stars

ORCA Agent Skills is a runtime that turns repeatable agent reasoning into executable, reusable skills. It solves the problem of logic trapped in prompts by providing a DAG-based execution layer with capabilities, bindings, and policy controls. It is for developers building agent systems who want portability, testability, and observability across different backends and model providers.

✨ Key features

  • Declarative DAG-based skill workflows
  • Backend-agnostic capability contracts
  • Multiple bindings: PythonCall, OpenAPI, MCP, OpenRPC
  • Policy and safety gates in runtime
  • CognitiveState and traceability for audit
  • Local deterministic baselines, no API key required

🎯 Use cases

  • Run a language summary skill locally without API keys
  • Execute decision workflows with structured outputs and confidence scores
  • Integrate skills into LangChain, CrewAI, AutoGen, or Semantic Kernel
  • Serve skills via HTTP API or MCP server for tool ecosystems
  • Author and package reusable cognitive workflows as DAGs

📦 Installation

🧰 Requirements: Python environment; no API key required for local runs; optional OpenAI API key for some integrations.

git clone https://github.com/gfernandf/agent-skills.git
cd agent-skills
make bootstrap

For Windows PowerShell:

git clone https://github.com/gfernandf/agent-skills.git
cd agent-skills
pip install -e ".[all,dev]"
git clone https://github.com/gfernandf/agent-skill-registry.git ../agent-skill-registry

🚀 Usage

python skills.py doctor
python skills.py run text.language-summary \
  --input '{"text": "ORCA turns agent reasoning into reusable executable skills."}'

⚠️ Good to know

The project is in active development; some features like auth, RBAC, and webhooks are advanced and may require additional setup. The README notes that the project status is documented in docs/PROJECT_STATUS.md.

❓ FAQ

Do I need an API key to run skills locally?

No, the local-first runs use deterministic Python baselines and do not require an API key.

What integration modes are available?

You can use the embedded SDK, HTTP API, MCP server, framework adapters (LangChain, CrewAI, AutoGen, Semantic Kernel), or native tool definitions for Anthropic, OpenAI, and Gemini.

How do I run a skill with the CLI?

Use python skills.py run <skill-name> --input '{"key": "value"}' after installing and running python skills.py doctor.

What is the difference between a skill and a capability?

A skill is a reusable cognitive workflow declared as a DAG, while a capability is a backend-agnostic contract with typed inputs and outputs.

📊 Repository

Stars★ 61
Forks🍴 5
Open issues🐛 86
Last commit🕒 Jul 1, 2026
Created📅 Mar 2026
Language💻 Python
License⚖️ Apache-2.0

🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.