A specialized protocol for AI agents to escalate complex blockers to a supervisor agent or human via webhook gateways.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-escalation
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-escalation 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 Escalation Skill?
The Escalation Skill is a critical reliability component within the Openclaw Skills ecosystem. It provides a structured mechanism for autonomous agents to recognize their own limitations and transfer context to a more capable supervisor entity. This ensures that technical blockers, configuration errors, or high-stakes decisions are handled with precision rather than resulting in infinite loops or hallucinated solutions.
By implementing this skill, developers can maintain high service levels for AI-driven workflows. It acts as a safety net that bridges the gap between autonomous task execution and human-in-the-loop oversight, utilizing a dedicated webhook gateway to maintain session integrity and complete problem context during the handoff.
Escalation Skill Use Cases
- Persistent technical errors or API failures that remain unresolved after 2-3 automated attempts.
- Scenarios requiring high-level permissions or infrastructure configurations outside the agent's scope.
- Ambiguous client requests that necessitate human judgment or strategic decision-making.
- Handling expired authentication tokens or security credentials that require manual re-authorization.
How Escalation Skill Works
- The AI agent identifies a persistent failure or out-of-scope request that meets the escalation criteria.
- The agent executes the local escalation script, passing structured arguments regarding the problem, previous attempts, and error logs.
- The script triggers a POST request to the
/hooks/agentendpoint on the local gateway. - A supervisor session is initialized, inheriting the full context of the previous interaction to prevent redundant work.
- The supervisor resolves the issue or further escalates the notification to a human operator via communication channels like Telegram.
Escalation Skill Setup
To integrate this capability into your agent environment using Openclaw Skills, ensure the script is accessible in your skill directory:
# Example of triggering an escalation manually or via agent exec
bash <SKILL_DIR>/scripts/escalate.sh \
"[Problem Description]" \
"[What was attempted]" \
"[Current error/result]"
Ensure your local gateway is configured to listen for incoming hooks at the designated endpoint.
Escalation Skill Data Schema & Taxonomy
The escalation process organizes data into three primary fields to ensure the supervisor has sufficient context:
| Field | Requirement | Description |
|---|---|---|
| Problem | Mandatory | A concise summary of the specific blocker or technical hurdle. |
| Attempts | Mandatory | A detailed log of what actions the agent took and why they failed. |
| Result | Optional | The final error message, stack trace, or state of the environment. |
All data is transmitted via JSON payload to the POST /hooks/agent endpoint.
Escalation Skill Advanced Features
- Human-in-the-loop integration for high-stakes decision-making environments.
- Multi-level escalation paths (e.g., Agent -> Supervisor -> Human Operator).
- Session persistence that allows the supervisor to see the exact state of the agent before the handoff.
- Automated Telegram notifications for real-time visibility into agent blockers within Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Summarize
A high-speed CLI tool designed to extract and summarize content from URLs, YouTube transcripts, and local documents using various LLM providers.
Border Radius Generator
A precise utility for generating CSS and Tailwind CSS border-radius code for uniform, per-corner, or complex geometric shapes.
JSON Formatter
A high-performance utility for prettifying, compressing, and validating JSON data structures with configurable indentation.
Marketing Designer
A senior-level creative strategist skill that bridges the gap between emotional human truths and data-driven marketing precision.
Aspect Ratio Calculator
A technical utility for calculating simplified aspect ratios, finding missing dimensions, and identifying industry-standard display formats.
Douyin & TikTok Trend Fetcher
An automated analytics tool that extracts real-time trending hashtags, viral music, and top creator data from TikTok and Douyin.
Featured*
