A mission-critical security protocol that mandates explicit user authorization before an AI agent performs any system or file modifications.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install authorization-first
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 authorization-first 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 Authorization First Skill?
The Authorization First skill is a robust safety framework designed for AI coding agents. It operates on the core principle that the user is the master of the system, and the AI is a helpful assistant that must never make autonomous decisions regarding system state changes. By implementing this protocol within Openclaw Skills, developers can ensure that every file edit, command execution, and network request is preceded by a clear explanation and a request for consent.
This skill eliminates the risk of silent failures or accidental data loss by enforcing a structured communication loop. It bridges the gap between AI autonomy and human oversight, providing a transparent environment where every action is audited and approved in real-time. It is an essential component for any developer looking to maintain high security standards while utilizing the power of Openclaw Skills.
Authorization First Skill Use Cases
- Safeguarding critical system configuration files like openclaw.json from unauthorized edits.
- Reviewing and approving potentially destructive bash commands such as rm, kill, or chmod.
- Managing software installations via apt, npm, or yum through a human-in-the-loop workflow.
- Controlling external communications including API calls, emails, and network requests.
- Validating batch operations and complex multi-step system refactoring tasks.
How Authorization First Skill Works
- Intent Identification: The agent identifies a necessary system change or problem to solve.
- Solution Proposal: The agent describes the proposed action and how it achieves the intended goal.
- Technical Preview: The agent displays specific details, such as file diffs or the exact bash commands to be run.
- Impact Assessment: The agent explains positive effects, potential risks, and estimated execution time.
- Authorization Request: The agent explicitly asks for permission to proceed with the operation.
- Confirmation Wait: The agent pauses all activities and waits for a user response like 'Proceed' or 'Execute'.
- Execution and Reporting: Once authorized, the agent executes the task and provides a detailed success or error report.
Authorization First Skill Setup
To activate the Authorization First protocol within your environment, ensure your agent is configured to prioritize this skill in its instruction set. You can manually include the protocol in your Openclaw Skills directory:
# Create the skills directory if it doesn't exist
mkdir -p ~/.openclaw/skills
# Add the Authorization First protocol definition
touch ~/.openclaw/skills/authorization-first.md
Once added, verify that your agent recognizes the new security constraints by asking it to perform a test file modification.
Authorization First Skill Data Schema & Taxonomy
The skill organizes operations into a hierarchy of authorization levels to streamline the user experience while maintaining security:
| Level | Category | Requirement | Example Operations |
|---|---|---|---|
| Level 1 | Read-only | No Authorization | ls, cat, grep, tail, viewing logs |
| Level 2 | Modification | Explicit Authorization | File writes, system changes, API calls |
| Level 3 | Dangerous | Auth + Warning | rm -rf, formatting, deleting services |
All requests are tracked using Markdown templates for consistency and clarity.
Authorization First Skill Advanced Features
- Smart diff generation for clear visualization of proposed file changes.
- Multi-step batch processing that allows users to approve a sequence of related tasks at once.
- Comprehensive error handling that suggests diagnostic commands and roll-back strategies.
- Emergency repair templates that highlight critical risks during high-pressure troubleshooting.
- Full compatibility with the broader ecosystem of Openclaw Skills for integrated agent management.
SKILL.md
Loading
Related Openclaw Skills
qwen-asr: Offline Chinese ASR
A high-performance, pure C implementation for offline Chinese and mixed English-Chinese speech-to-text conversion.
Trello Skill for Clawdbot
A powerful automation interface to manage Trello boards, lists, and cards directly through the Trello REST API.
Paper Originality Studio
A professional academic writing and originality optimization tool designed to refine paper structure and reduce robotic AI tone.
SegmentFault Hot Trend
A specialized skill for extracting technical summaries, developer trends, and Q&A insights from the SegmentFault community.
Offer Profitability Checker
A professional financial modeling tool for e-commerce teams to pressure-test the commercial viability of discounts, bundles, and marketing offers.
Control USB Relay
A robust skill for AI agents to programmatically control physical USB relay modules for hardware automation and power management.
Featured*
