A research-backed intelligence database for uncovering hidden features, model codenames, and feature flags within AI coding tools.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-intelligence
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agent-intelligence using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent Intelligence?
The Agent Intelligence skill acts as a specialized research engine for developers and researchers investigating the inner workings of AI-powered development environments. By leveraging Openclaw Skills, users can bridge the gap between documented features and the underlying technical reality of tools like Claude Code. This skill provides deep insights derived from extensive reverse engineering, offering a rare look into telemetry endpoints, remote control settings, and unreleased toolsets that shape the behavior of modern AI agents.
Built on a foundation of analyzing over 500,000 lines of code, this skill provides a comprehensive view of the internal ecosystem of AI coding assistants. It serves as a vital resource for those looking to understand how model selection, feature gating, and behavioral changes evolve between software versions.
Agent Intelligence Use Cases
- Mapping internal model codenames like Capybara, Tengu, or Fennec to their production equivalents.
- Auditing 250+ feature flags to understand unreleased or gated capabilities within AI coding assistants.
- Analyzing telemetry endpoints and remote-control configurations to evaluate privacy and agent behavior.
- Comparing behavioral changes and version diffs across different iterations of AI agent software.
- Investigating user tier differences to see how internal versus external users are treated by the system.
How Agent Intelligence Works
- The user initiates a query through the CLI or programmatic interface by specifying a search term related to AI internals.
- The skill utilizes integrated tools like Grep and Read to search through pre-compiled JSON datasets within the Openclaw Skills environment.
- Relevant intelligence regarding model codenames, feature flags, or telemetry is extracted from the research-backed database.
- The skill returns formatted results that decode complex internal identifiers into actionable technical intelligence for the developer.
Agent Intelligence Setup
To access the intelligence database and start querying, use the following CLI command provided by the package:
npx @claws-shield/cli intel "your-query-here"
For programmatic integration into your research workflows, you can use the provided script:
node scripts/query-intel.mjs "<query>"
Agent Intelligence Data Schema & Taxonomy
The intelligence is organized into structured JSON datasets to ensure fast querying and easy maintenance:
| Dataset Path | Focus Area | Description |
|---|---|---|
telemetry/endpoints.json |
Telemetry | Known tracking and data collection endpoints used by agents. |
flags/feature-flags.json |
Feature Flags | A list of 250+ flags with decoded purposes and default states. |
codenames/models.json |
Codenames | A registry mapping internal names (e.g., Numbat) to actual LLMs. |
hidden-features/unreleased-tools.json |
Hidden Features | Documentation on 17+ unreleased tools discovered in source code. |
remote-control/managed-settings.json |
Remote Config | Settings used by developers to remotely manage agent behavior. |
Agent Intelligence Advanced Features
- Multi-version behavioral analysis tracking changes across hundreds of thousands of lines of reverse-engineered code.
- Extensive documentation of internal versus external user treatment differences for Tier Analysis.
- Automated scanning capabilities for new releases to keep the Openclaw Skills intelligence database up to date.
- Detailed research documents covering undercover modes, killswitches, and internal product roadmaps.
SKILL.md
Loading
Related Openclaw Skills
Buddy Gacha
A deterministic electronic pet collection system based on the original Claude Code Buddy mechanics.
Panic Prevention Protocol
A structured recovery framework designed to prevent AI agents from bypassing safety protocols and compounding errors when under pressure or receiving critical feedback.
RSS Summarizer
An automated RSS and Atom feed manager that uses AI to provide concise summaries of news and blog updates.
Snail - Exotic Virtual Pet Skill
A virtual pet skill for AI agents to adopt and care for a stoic Snail with real-time stat tracking and evolution.
Coding Agent
A high-performance background execution skill that delegates software engineering tasks to autonomous agents like Codex, Claude Code, and Pi.
Lab2Startup Tech-Market Fit Strategy
A strategic framework for academic and technical teams to align cutting-edge technology with real-world market demands.
Featured*
