Skills & Modular Capabilities

ClawPaw performs specific categories of actions through modular execution units called Skills.

These capabilities are not automatically orchestrated in the background. They are explicitly selected or invoked within a task.

You can browse, search, and enable abilities through the Abilities panel, or use @ syntax in the input field to call a specific Skill directly.

Capabilities are visible, selectable, and controllable.

What Are Skills?

Skills are structured execution modules responsible for specific types of tasks. They allow ClawPaw to interact with your system within clearly defined boundaries, rather than granting unrestricted access.

Common skill domains include:

• File inspection and organization
• Document parsing and summarization
• Browser content reading
• Calendar information retrieval
• Structured content generation

Each Skill operates under defined constraints:

• Access limited to approved folders
• Governed by macOS permission controls
• Bound by the selected autonomy mode
• Executed under a sandbox-first policy

Skills do not have system-wide authority. They only operate when explicitly invoked.

How to Use Skills

There are two ways to use Skills:

1. Through the Abilities Panel

• Open Abilities
• Search or browse available Skills
• Select the capability you want to use

For example, selecting Find Skills enters discovery mode.

2. Using @ Commands

You can explicitly invoke a Skill in the input field:

@Skills:find-skills

@Skills:skill-creator

This ensures the specified capability is used for the current task.

Find Skills: Capability Discovery

Selecting Find Skills from the Abilities panel, or invoking it via @Skills:find-skills, enters capability discovery mode.

In this mode, you can:

• Search existing Skills
• Browse supported capability categories
• Understand what each Skill is designed to do

This is useful for exploring system capabilities or intentionally choosing a specific execution path.

Skill Creator Guide: Creating Capabilities

ClawPaw also supports a Skill Creator Guide mode.

By invoking:

@Skills:skill-creator

you enter a structured skill design flow.

This mode allows you to:

• Design new skill structures
• Define capability boundaries
• Specify input and output formats
• Establish execution behavior

Skill Creator Guide provides a structured capability design process. It does not grant direct system-level authority.

Even custom-created Skills remain constrained by:

• Approved folder scope
• macOS permission limits
• The active autonomy mode
• Sandbox execution boundaries

Capabilities can expand. Boundaries remain intact.

---

MCP Modules

MCP (Modular Capability Protocol) modules support more structured or domain-specific workflows.

MCP characteristics include:

• Task-scoped execution
• No persistent global authority
• No ability to override permission boundaries
• Operation within sandbox and approved scopes

MCP extends capability depth without altering the core security model.

Design Principle: Explicit Capability, Clear Boundaries

ClawPaw’s capability system is built on two principles:

• Capabilities must be invoked
• Authority remains bounded

Skills and MCP modules cannot:

• Access unapproved directories
• Escalate privileges
• Bypass macOS privacy controls
• Run hidden background processes

Capabilities are modular — and constrained by design.

ClawPaw allows you to compose abilities through conversation, but execution always occurs within visible, user-defined boundaries.