finclip-openskills 0.4.0

OpenSkill Runtime Python bindings - Claude Skills compatible runtime with WASM sandbox

OpenSkills Python Bindings

Python bindings for OpenSkills Runtime - Claude Skills compatible runtime with WASM sandbox.

Installation

pip install finclip-openskills

Note: Pre-built wheels are available for macOS and Linux only. Windows users need to build from source:

git clone https://github.com/Geeksfino/openskills.git
cd openskills/bindings/python
pip install maturin
maturin develop

Usage

Basic Runtime API

from openskills import OpenSkillRuntime

# Create runtime (discovers from standard locations)
runtime = OpenSkillRuntime()

# Or specify project root
runtime = OpenSkillRuntime.with_project_root('/path/to/project')

# Or load from specific directory
runtime = OpenSkillRuntime.from_directory('/path/to/skills')

# Discover skills from standard locations
skills = runtime.discover_skills()
print(f"Found {len(skills)} skills")

# List skills (progressive disclosure)
for skill in runtime.list_skills():
    print(f"{skill['id']}: {skill['description']}")

# Activate a skill (load full content)
loaded = runtime.activate_skill('my-skill')
print(loaded['instructions'])

# Execute WASM module
result = runtime.execute_skill(
    'my-skill',
    input={'query': 'hello'},
    timeout_ms=5000
)

print(result['output'])
print(result['audit'])

# Check tool permissions
can_read = runtime.is_tool_allowed('my-skill', 'Read')

⭐ Pre-built Tools (Recommended)

For agent integration, use pre-built tools that eliminate boilerplate:

LangChain Integration:

from openskills import OpenSkillRuntime
from openskills_tools import create_langchain_tools, get_agent_system_prompt

# Initialize runtime
runtime = OpenSkillRuntime.from_directory('./skills')
runtime.discover_skills()

# Create LangChain-compatible tools
tools = create_langchain_tools(runtime, workspace_dir='./output')

# Get skill-agnostic system prompt
system_prompt = get_agent_system_prompt(runtime)

# Use with LangChain
from langchain.agents import create_agent
agent = create_agent(model, tools, system_prompt=system_prompt)

Framework-Agnostic (Simple Functions):

from openskills import OpenSkillRuntime
from openskills_tools import create_simple_tools

runtime = OpenSkillRuntime.from_directory('./skills')
runtime.discover_skills()

# Create simple callable functions (works with any framework)
tools = create_simple_tools(runtime, workspace_dir='./output')

# Use tools directly
skills = tools['list_skills']()
loaded = tools['activate_skill']('my-skill')
tools['write_file']('output.txt', 'Hello, World!')

# Run a WASM module from a skill (auto-detected by file extension)
result = tools['run_skill_script'](
    'skill-creator',
    'wasm/skill.wasm',
    input={'action': 'init_skill', 'skill_name': 'my-skill', 'path': 'skills/public'}
)
print(result['output'])

Available Tools:

• list_skills - List available skills
• activate_skill - Load full SKILL.md instructions
• read_skill_file - Read helper files from skills
• list_skill_files - List files in skill directories
• run_skill_script - Execute sandboxed scripts or WASM modules
• write_file - Write to workspace (with security validation)
• read_file - Read from workspace (with security validation)
• list_workspace_files - List files in workspace
• get_file_info - Get file metadata

Benefits:

• ✅ ~200 lines less code: No manual tool definitions
• ✅ Security built-in: Path validation, workspace isolation
• ✅ Workspace management: Automatic sandboxed file I/O
• ✅ Skill-agnostic: Works with any skill without code changes