openclaw-agent-compute_skill

This skill helps you securely invoke compute.* tools through a private gateway, enabling remote execution and usage tracking via HTTP.
  • Python

2.6k

GitHub Stars

4

Bundled Files

2 months ago

Catalog Refreshed

3 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill openclaw/skills --skill openclaw-agent-compute

  • _meta.json485 B
  • package.json416 B
  • README.md1.1 KB
  • SKILL.md1.7 KB

Overview

This skill is a public HTTP client that exposes compute.* tools by proxying requests to a private Compute Gateway over HTTPS. It provides a minimal agent-friendly interface plus a starter kit that runs OpenClaw preconfigured. The package includes example scripts and a local smoke test to validate connectivity and basic execution.

How this skill works

The skill calls a private Compute Gateway using environment-configured endpoints and an API key. It implements session lifecycle and execution flows by hitting /v1/sessions to create sessions, /v1/exec to run commands, /v1/usage/{session_id} to fetch usage and /v1/sessions/{session_id} to destroy sessions. Example scripts show how to run the HTTP client and an example execution flow locally.

When to use it

  • You need an agent-facing wrapper to run compute.* tools through a centralized, private gateway.
  • You want to run OpenClaw preconfigured locally or inside a controlled environment.
  • You must separate public agent interfaces from private compute infrastructure for security or compliance.
  • You need to track usage/cost per session via the gateway's usage endpoint.
  • You want a lightweight starter kit to bootstrap OpenClaw without building a full image pipeline.

Best practices

  • Store MCP_COMPUTE_URL and MCP_COMPUTE_API_KEY securely in environment or a secrets manager.
  • Run the provided local smoke test after updating environment variables to verify connectivity.
  • Limit API key scope and rotate keys regularly to reduce blast radius.
  • Use the usage endpoint to monitor cost and enforce session quotas.
  • Override OPENCLAW_IMAGE only when you control the image lifecycle and trust the tag/source.

Example use cases

  • Expose compute.* capabilities to agents while keeping computation behind a private gateway.
  • Bootstrap a developer environment that runs OpenClaw using the starter kit and an overrideable image.
  • Integrate session-based execution into a larger orchestration system that needs per-session accounting.
  • Quickly validate gateway configuration and credentials with the included example_exec script.
  • Archive or back up skill versions and scripts while preserving a consistent HTTP client interface.

FAQ

You must set MCP_COMPUTE_URL (the gateway HTTPS URL) and MCP_COMPUTE_API_KEY.

How do I test the client locally?

Copy the .env example, update values, run npm install, then run the provided example: npm run example:exec.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
openclaw-agent-compute skill by openclaw/skills | VeilStrat