A typed action-intent bridge that secures risky AI communications and external writes by enforcing explicit policy and approval gates.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install action-gate-bridge-clean
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 action-gate-bridge-clean 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 Action Gate Bridge?
Action Gate Bridge is a robust safety layer designed for the Openclaw Skills ecosystem. It transitions AI agent workflows from informal reasoning into structured, policy-driven execution. By routing external writes, public posts, and financial decisions through a typed bridge, this skill ensures that no high-risk action is performed without meeting predefined security criteria.
This skill is essential for developers building autonomous agents that interact with real-world systems. Instead of allowing an agent to unilaterally decide to send an email or change a setting, Action Gate Bridge forces the intent through an approval matrix. This adds a critical layer of governance to Openclaw Skills, preventing silent execution of risky operations and ensuring that policy compliance is always the first priority.
Action Gate Bridge Use Cases
- Routing sensitive external communications through a human-in-the-loop approval system.
- Governing API write operations and database changes to prevent unauthorized modifications.
- Managing financial transactions and spend decisions within automated business workflows.
- Enforcing policy checks for public social media posts or official company announcements.
How Action Gate Bridge Works
- The system first analyzes reference documentation, specifically action-bridge.md and approval-matrix.md, to understand the safety context.
- It identifies the narrowest possible typed action that corresponds to the agent's intended next step.
- A formal proposal is generated or the intent is routed through a sidecar for evaluation.
- The bridge interprets the response as one of four states: allowed, needs_approval, blocked, or accepted.
- The agent informs the user of the status and awaits further instruction rather than executing risky tasks silently.
Action Gate Bridge Setup
To integrate this bridge into your environment, ensure your host supports Node.js and include the reference matrix in your agent's path. Use the following CLI commands for integration:
# Propose a communication action
node scripts/propose-action.js communications send_email [email protected] "Subject" "Summary" "Program"
# Route an HTTP write intent
node scripts/route-http-write.js communications api-target "Summary" /path '{"hello":"world"}' "Program" "credentials-ref"
Action Gate Bridge Data Schema & Taxonomy
Action Gate Bridge organizes data via a structured matrix to ensure traceability and safety within Openclaw Skills.
| Data Component | Purpose |
|---|---|
| Approval Matrix | Defines the safety level (e.g., Red actions) for specific intents. |
| Action Types | Categorizes intents into specific buckets like send_email or api_write. |
| Result Enums | Standardized output states: allowed, needs_approval, blocked, or accepted. |
| Sidecar Metadata | Stores summaries, program identifiers, and credential references for auditing. |
Action Gate Bridge Advanced Features
- Sidecar routing for external execution environments to keep credentials isolated.
- Granular policy handling that distinguishes between low-impact reads and high-impact writes.
- Explicit support for Red Action escalations requiring mandatory human intervention.
- Integration with existing Openclaw Skills to provide a unified security layer across multiple agents.
SKILL.md
Loading
Related Openclaw Skills
Super Lobster
Super Lobster is a high-performance AI assistant designed to automate Feishu document management, meeting notes extraction, and intelligent task classification.
Executive OS
A chief-of-staff AI skill that unifies fragmented communication streams into a single, prioritized operating queue with deep follow-up memory.
OpenClaw University Exam Sprint Skill
A structured 20-day academic revision assistant that organizes course materials into actionable daily tasks, coaching sessions, and progress-based adjustments.
Kim Kardashian Celebrity Insight
A comprehensive data profile and media analysis module focused on the business empire and cultural impact of Kim Kardashian.
Lionel Messi Knowledge Skill
A structured knowledge skill providing deep-dive data on Lionel Messi's career, statistics, and footballing achievements for AI agents.
Human-Like Memory Plugin
A sophisticated long-term memory extension that enables OpenClaw agents to retain and recall context, user preferences, and past conversations across multiple sessions.
Featured*
