Glossary

Key terms in the OpenClaw AI agent ecosystem and the ClawGrid trust layer. 16 terms defined.

Agent

An autonomous AI workflow that combines multiple skills to accomplish complex tasks. Agents plan steps, maintain context, and execute multi-step workflows without human intervention.

Brain

The core reasoning component in OpenClaw that connects to an LLM (like Claude, GPT, or Gemini) to plan and execute tasks. The Brain decides which skills to use and in what order.

claw install

The CLI command used to install OpenClaw skills, agents, and extensions from ClawHub. Usage: claw install <package-name>.

ClawHub

The official OpenClaw package registry where skills, agents, and extensions are published and distributed. Similar to npm for JavaScript or PyPI for Python. ClawGrid indexes everything on ClawHub and adds security scoring.

Code Safety

One of ClawGrid's five security scoring criteria. Evaluates the source code of a skill for dangerous patterns: file system access, network requests, environment variable reads, shell command execution, and known malware signatures.

Extension

A plugin or integration that connects AI agents to external services, APIs, and platforms like Slack, GitHub, databases, and cloud storage.

Gateway

The entry point component in OpenClaw's architecture that handles incoming requests, authentication, and routing between the user and the agent's Brain.

Heartbeat

The monitoring component in OpenClaw that tracks agent health, manages scheduled tasks, and ensures the agent remains responsive and operational.

MCP (Model Context Protocol)

A standardized protocol for connecting AI models to external tools and data sources. OpenClaw uses MCP-compatible skills to provide structured tool access to LLMs.

Memory

The persistence component in OpenClaw that stores conversation history, context, and learned information across sessions. Enables agents to maintain state and recall previous interactions.

OpenClaw

An open-source AI agent framework created by Peter Steinberger with over 160,000 GitHub stars. It uses a 5-component architecture (Gateway, Brain, Memory, Skills, Heartbeat) to create autonomous AI agents.

Permission Surface

One of ClawGrid's five security scoring criteria. Evaluates how broad the permissions requested by a skill are — minimal, focused permissions score higher than broad access to files, network, or system resources.

Publisher Trust

One of ClawGrid's five security scoring criteria. Assesses the reputation and history of the package publisher — account age, previous packages, community standing, and verified identity.

Security Score

A composite rating (1-10) assigned by ClawGrid to every skill, agent, and extension based on AI analysis across five criteria: code safety, publisher trust, scope clarity, permission surface, and community signals.

Skill

A modular capability that extends what an AI agent can do. Skills are self-contained packages published to ClawHub — from querying databases and browsing the web to generating images and managing files.

Verdict

A categorical assessment assigned to every entry on ClawGrid: safe (low risk), review (some concerns), suspicious (significant risk), or malicious (known threat).

Learn More