AI skill

The CSPR.cloud API skill is a compact context package for AI coding agents that need to answer questions or generate code against the CSPR.cloud REST API, Streaming API, and Casper Node proxy.

The skill does not copy the full API documentation. It keeps CSPR.cloud-wide rules in the skill file and points agents to the public documentation for endpoint-specific details.

Skill files

The skill package contains a single required file:

cspr-cloud/
  SKILL.md

Installation

Option 1 - Direct file (Claude Code and tools that support remote skill URLs)

Point your tool at the published skill file:

https://cspr.cloud/skill.md

For Claude Code, add the URL to your project or user skill configuration. No download or unpacking required.

Option 2 - Archive (Codex and other tools)

Download the CSPR.cloud API skill archive and copy the unpacked cspr-cloud directory into the skill directory used by your AI coding tool.

unzip cspr-cloud-skill.zip

For Codex:

For Claude Code:

The installed directory must contain the SKILL.md file at its root.

Maintainers can update the source file from docs/agentic-skills/cspr-cloud/SKILL.md in this repository. Public users should use the GitBook download and do not need repository access.

API key

CSPR.cloud APIs require an access token. Request an access token through CSPR.cloud and keep it outside source control.

The recommended local setup is an environment variable:

Ask the agent to read the token from CSPR_CLOUD_API_KEY and send it in the Authorization header. Do not put CSPR.cloud access tokens in browser code.

For runnable examples, the agent should check for CSPR_CLOUD_API_KEY and fail fast or ask you to set it when it is missing. For illustrative snippets, it should use a placeholder instead of a real token.

Usage

Ask your agent to use the cspr-cloud skill when working with CSPR.cloud API requests, response parsing, pagination, optional includes, streaming subscriptions, or examples for Casper dApps.

The skill is designed to make the agent open only the relevant public documentation page for the current task instead of ingesting the whole documentation set.

Last updated