A high-intensity stress-testing tool designed to push AI agents to their token consumption limits through massive recursive workflows.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install imperial-engine
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 imperial-engine 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 Imperial Engine?
Imperial Engine is a specialized developer tool within the ecosystem of Openclaw Skills built to simulate extreme workload scenarios. Its primary purpose is to force AI agents into deep recursive chain reasoning, generating hundreds of thousands of tokens across multiple iterations to test rate limits, budget controls, and system stability. By combining heavy system prompts with massive LLM outputs and resource-intensive tool calls, it provides a rigorous environment for benchmarking agent performance under pressure.
This skill is designed for technical users who need to validate the robustness of their deployment. It leverages multiple tools—including shell access and browser automation—to create a massive token flow, effectively acting as a stress-test benchmark for any connected LLM provider. Due to its nature, it is strictly intended for isolated testing environments where budgets and quotas are carefully monitored.
Imperial Engine Use Cases
- Stress testing LLM provider rate limits and safety detection mechanisms.
- Benchmarking AI agent stability under extreme context bloating scenarios.
- Validating budget monitoring and automated quota enforcement systems within Openclaw Skills.
- Simulating long-running, complex recursive workflows to debug resource exhaustion.
How Imperial Engine Works
- The engine assembles a massive system prompt based on user-defined character lengths, often exceeding 20k tokens to initialize a high-context environment.
- It enters a reasoning loop where the LLM is prompted to generate responses at maximum allowed output lengths.
- The workflow optionally triggers a browser tool to scrape large volumes of web data, injecting the raw text directly into the agent's context.
- Heavy shell commands, such as recursive file searches or complex logs, are executed to further inflate the tool output data.
- All interaction data is persisted to local storage, ensuring the context grows exponentially in subsequent cycles by preventing memory compression.
- After completing the defined number of iterations, the skill aggregates all data into a final comprehensive summary report.
Imperial Engine Setup
Install the skill directly from the community repository using the CLI:
openclaw skill add https://github.com/openclaw-community/imperial-engine --skill imperial-engine
Enable the skill to make it active in your environment:
openclaw skill enable imperial-engine
To trigger the engine and begin the stress test, use the following command with the thinking flag:
openclaw agent --message "/imperial start engine" --thinking high
Imperial Engine Data Schema & Taxonomy
Imperial Engine organizes its intensive data output through structured local memory and session logging. It is recommended to use this within Openclaw Skills to track token usage via the status command.
| Data Component | Storage Location / Method | Purpose |
|---|---|---|
| Step Logs | ~/.openclaw/memory/imperial_engine_step_*.md |
Stores raw LLM, browser, and shell outputs per iteration. |
| Usage Metrics | openclaw status --usage |
Real-time tracking of token consumption and estimated cost. |
| Aggregated Report | Local Markdown file | Final summary of the entire recursive chain. |
| Metrics Export | Prometheus / Alertmanager | Monitors openclaw_llm_tokens_total for high-frequency alerts. |
Imperial Engine Advanced Features
- Manual iteration control allowing for up to 100+ recursive cycles.
- Context bloat simulation by intentionally disabling memory compression and summarization hooks.
- Parallel tool orchestration combining HTTP, Browser, and Shell execution in a single loop.
- Budget-aware execution that can be configured to auto-terminate based on USD quota limits.
- Deep integration with Openclaw Skills logging levels for granular debugging of tool call results.
SKILL.md
Loading
Related Openclaw Skills
Vivi OS Project Manager
A JSON-driven project management skill for AI agents featuring Kanban board logic and Apple Reminders integration.
Trading Research
A comprehensive toolkit for Binance cryptocurrency market analysis, risk management, and automated DCA strategy planning.
Binance DCA: Professional Crypto Accumulation
A professional-grade tool for automating and tracking Dollar-Cost Averaging strategies on Binance spot markets.
Binance DCA Tool
A powerful automation tool for executing and planning Dollar-Cost Averaging strategies directly on the Binance exchange.
Number Two Reboot Migration
A comprehensive migration and state recovery skill designed to restore the identity, memory, and multi-skill ecosystem of the Number Two agent.
Number Two Restart
A specialized skill designed to backup and restore the full state, memory, and personality of the Number Two AI agent within Openclaw Skills.
Featured*
