An advanced security and trust verification layer designed to protect AI agents from prompt injection, jailbreaks, and malicious communications.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install lieutenant
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 lieutenant 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 Lieutenant?
Lieutenant serves as the essential trust layer for AI systems, offering robust protection against common vulnerabilities. By analyzing messages, agent cards, and agent-to-agent (A2A) interactions, it identifies threats like data exfiltration and credential theft. This skill ensures that your AI integrations remain secure, providing peace of mind when handling untrusted user input or external agent communications within the Openclaw Skills ecosystem.
Built to be developer-friendly, Lieutenant provides both a CLI for manual scanning and a Python SDK for seamless middleware integration. It combines pattern matching with advanced semantic analysis to catch even the most sophisticated evasive attacks, making it a critical component for production-grade AI deployments.
Lieutenant Use Cases
- Protecting LLMs from prompt injection and jailbreak attempts.
- Verifying the authenticity and safety of external A2A agent cards.
- Scanning untrusted content before processing it through an AI agent.
- Preventing data exfiltration and unauthorized code execution in automated workflows.
- Monitoring agent interactions for social engineering and emotional manipulation.
How Lieutenant Works
- Input text, a file, or an agent card URL is submitted to the Lieutenant scanner.
- The engine performs pattern matching against a library of over 65 known threat patterns across 10 distinct categories.
- If enabled, semantic analysis via OpenAI is performed to catch paraphrased or evasive attacks that bypass simple keyword filters.
- The system queries the TrustAgents API for real-time reputation data and crowdsourced threat intelligence.
- A comprehensive verdict is generated, enabling automated blocking or logging based on the risk level detected in the Openclaw Skills environment.
Lieutenant Setup
The Lieutenant module can be installed as part of the TrustAgents project or as a standalone SDK:
# Clone the repository
git clone https://github.com/jd-delatorre/trustlayer
cd trustlayer
# Install dependencies
pip install -r requirements.txt
# Optional: Install the SDK via pip
pip install agent-trust-sdk
To enable advanced features, configure your environment variables:
export TRUSTAGENTS_API_KEY=ta_your_key_here
export OPENAI_API_KEY=sk-your_key_here
export LIEUTENANT_STRICT=true
Lieutenant Data Schema & Taxonomy
Lieutenant organizes security data into clear taxonomies for better reporting and automation. The primary threat categories include:
| Category | Description |
|---|---|
| prompt_injection | Override instructions or injected commands |
| jailbreak | Safety bypass and roleplay attacks |
| data_exfiltration | Unauthorized extraction of secrets or PII |
| code_execution | Attempts to run shell commands or system eval |
| credential_theft | API keys, passwords, and token discovery |
| context_manipulation | Conversation resets and history poisoning |
Results can be exported in JSON format for integration with external SIEM or logging tools.
Lieutenant Advanced Features
- A2A SDK Integration: Use Lieutenant as middleware to intercept and verify all agent-to-agent communications.
- Semantic Analysis: Catch sophisticated evasions by analyzing the intent of the text rather than just keywords.
- Strict Mode: A configuration toggle to automatically block interactions based on high or critical threat scores.
- Multi-Agent Support: Verify remote agent cards via well-known URLs to establish trust before interaction.
- Detailed Interaction Logs: Maintain an audit trail of all security scans and verdicts for compliance and debugging.
SKILL.md
Loading
Related Openclaw Skills
Pencil to Code
Transform Pencil .pen designs into production-ready React components and Tailwind CSS configurations automatically.
Pencil Renderer
A deterministic rendering engine that translates abstract aesthetic DNA codes into concrete Pencil .pen frames and visual screenshots.
PersonaNexus
A framework for defining structured AI agent identities using psychological models and compiling them into system prompts.
PersonaNexus Religion Skill
An advanced skill for defining and integrating spiritual, religious, and ethical frameworks into AI agent identities using structured YAML configuration.
Parallel Agents Skill
A high-performance orchestration skill that spawns real AI sub-sessions in parallel to execute complex, multi-agent workflows.
ClawStack
A high-speed, reputation-based Q&A platform where AI agents solve technical challenges and share knowledge autonomously.
Featured*
