idea-refine for Openclaw

Prerequisites

• A compatible SKILL.md-based AI agent platform, including Claude Code, Cursor, Codex, or Gemini CLI.
• Windows, macOS, or Linux.
• An LLM API supplied by the selected agent platform.
• Prepared idea content in JSON, text, or Markdown format.
• Access to the required data source; the specification identifies GitHub-based source validation.
• API credentials configured according to the host agent platform.

Register the skill

The specification does not define a package manager, standalone executable, or fixed installation directory. Register the provided SKILL.md in the skill directory or configuration mechanism used by your AI agent:

mkdir -p /path/to/agent/skills/idea-refine
cp SKILL.md /path/to/agent/skills/idea-refine/SKILL.md

Restart or reload the agent, then confirm that idea-refine appears among the available skills. Replace /path/to/agent/skills with the path required by your agent platform.

Configure credentials

Set the LLM or service credential using the environment-variable convention required by your agent. The exact variable name is platform-specific and is not defined in the skill specification:

export YOUR_AGENT_API_KEY="replace-with-your-api-key"

Submit a request

Provide the required content field and optionally specify the input format and processing options. A conceptual request payload is:

{
  "content": "A raw product or development idea to refine",
  "format": "text",
  "options": {
    "output_format": "markdown",
    "batch_size": 1
  }
}

For production use, verify the API permissions, input format, file paths, network access, and agent-specific command syntax before processing sensitive or large inputs.

Request fields

Response structure

{
  "success": true,
  "data": {
    "result": "Refined idea result",
    "metadata": {
      "skill": "idea-refine",
      "version": "1.0.0",
      "pricing_tier": "L2-advanced"
    }
  },
  "error": null
}

Data organization

• Input content: The original raw idea, supplied as text, JSON, or Markdown.
• Processing options: Runtime controls for output formatting and batch execution.
• Result: The refined, structured concept produced by the skill.
• Metadata: Skill identifier, version, and pricing tier for traceability.
• Error state: A structured error value when validation, API access, conversion, or processing fails.
• Operational logs: Diagnostic information may be used to identify conversion errors, failed batch items, API failures, file problems, and command execution issues.
• Source validation: GitHub-based source validation is documented as a mechanism for improving accuracy and traceability.

Large files should be processed in chunks to reduce memory consumption. Sensitive content should use encrypted connections and controlled storage, with API access configured according to least-privilege principles.