An advanced security and memory pipeline for AI agents that ensures safe plan execution through multi-dimensional guarding and evolutionary learning.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mia-trust
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 mia-trust 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 MIA-Trust Pipeline?
MIA-Trust (Memory-Intelligent Assistant Trust) is a sophisticated architectural pipeline designed to act as both a security gateway and a memory evolution layer for AI agents. It provides a robust framework for validating user queries, generating structured plans, and auditing those plans before execution to ensure maximum reliability. By integrating this component into Openclaw Skills, developers can create more resilient agents that learn from past interactions.
The system utilizes a dual-memory approach, combining standard execution logs with trust-based experience data. This allows the assistant to evolve over time, improving its decision-making capabilities and security posture through a continuous feedback loop and rigorous safety checks across multiple dimensions.
MIA-Trust Pipeline Use Cases
- Protecting AI agents from malicious prompts or unsafe query execution through six-dimensional pre-checks.
- Auditing complex execution plans in high-stakes environments before any actions are performed.
- Implementing long-term memory for AI agents to recall successful workflows and previous solutions.
- Establishing a feedback loop to refine AI behavior based on explicit user ratings and performance history.
How MIA-Trust Pipeline Works
- The system performs an initial safety pre-check (guard_blocked) across six distinct security dimensions to filter the user query.
- A specialized Planner module generates a detailed execution draft based on the validated query.
- The evaluate_plan module performs a rigorous three-round security audit on the proposed plan draft.
- Once validated, the plan is executed within the target environment.
- Execution results, feedback, and experiences are persisted to dual-memory storage (memory and trust_experience) for future retrieval and system evolution.
MIA-Trust Pipeline Setup
Navigate to the skill directory and install the necessary Node.js dependencies:
cd skills/mia-trust
npm install
Configure the planner by setting the required environment variables for your LLM provider:
export MIA_PLANNER_MODE=api
export MIA_PLANNER_API_KEY=your-api-key
export MIA_PLANNER_URL=https://your-api-endpoint/v1/chat/completions
export MIA_PLANNER_MODEL=your-model
You can then run the full pipeline with a single command:
./run.mjs "Your query here"
MIA-Trust Pipeline Data Schema & Taxonomy
MIA-Trust organizes its data using specialized flat files to ensure portability and high-speed access for Openclaw Skills agents:
| Data Component | File Path | Description |
|---|---|---|
| Execution Memory | memory/memory.jsonl |
Stores step-by-step history of questions, plans, and executions. |
| User Feedback | feedback/feedback.jsonl |
Records user ratings (e.g., 'good') and qualitative feedback for specific answers. |
| Trust Experience | trust/trust_experience.json |
An evolutionary dataset used to improve the plan evaluation logic over time. |
MIA-Trust Pipeline Advanced Features
- Multi-dimensional safety guarding that inspects queries across six security vectors.
- Three-round recursive plan evaluation to catch logic errors or security vulnerabilities before execution.
- Dual-layer memory architecture that separates operational logs from trust-based evolutionary data.
- Granular CLI tools for manual memory management, including searching, listing, and storing experiences.
- Seamless modular integration designed specifically for the Openclaw Skills ecosystem.
SKILL.md
Loading
Related Openclaw Skills
Stock Roundtable
A sophisticated investment research skill that simulates expert multi-perspective debates to provide evidence-based stock analysis and decision support.
Video Monetization Pro
A comprehensive end-to-end automation skill for AI video creators to research, produce, publish, and track revenue automatically.
Personal Fashion Assistant
An intelligent fashion consultant that manages your digital wardrobe, analyzes personal style metrics, and generates daily outfit recommendations.
Document Signing Setup Assistant
An expert guide for configuring professional e-signature workflows, from envelope creation to signer field mapping and status tracking.
Slack Assistant
A powerful command-line integration for Slack that leverages AI to summarize channels, draft responses, and manage workspace communication efficiently.
ClawHub Security Scan
A pre-publish security auditing tool designed to identify and fix code patterns that trigger security flags on ClawHub.
Featured*
