ClearPoint Nexus

Implementation Guides

Install skills in under 2 minutes

Step-by-step setup for each supported platform. Every skill ships with adapters for all three — just copy the right file and go.

🟠

Claude Code

File: SKILL.md

Setup Steps

  1. Download and unzip your skill package.
  2. Open the `claude-code/` directory inside the zip.
  3. Copy `SKILL.md` into your project root or working directory.
  4. Claude Code reads SKILL.md automatically when it starts a session in that directory.
  5. Start a conversation with Claude Code — the skill instructions are loaded as context.
  6. Ask Claude to perform the task described in the skill (e.g., "Run a deep research report on X").

Tips

  • SKILL.md works like a contract — Claude follows the structure, constraints, and output format defined in the file.
  • You can have multiple SKILL.md files in nested directories for different tasks.
  • The skill works with any Claude model (Sonnet, Opus, Haiku).
🟢

Codex

File: AGENTS.md

Setup Steps

  1. Download and unzip your skill package.
  2. Open the `codex/` directory inside the zip.
  3. Copy `AGENTS.md` into your project root (same level as package.json or .git).
  4. Codex reads AGENTS.md automatically when you run `codex exec` in that directory.
  5. Run your task: `codex exec "Perform deep research on X"`
  6. Codex follows the instructions, process, and rules defined in AGENTS.md.

Tips

  • AGENTS.md is an open standard — it also works with Jules, Cursor, Aider, Gemini CLI, and other compatible agents.
  • You can have a global AGENTS.md in `~/.codex/AGENTS.md` for cross-project defaults.
  • Nested AGENTS.md files override parent instructions for specific subdirectories.
🔵

OpenClaw

File: manifest.yaml + playbook.md

Setup Steps

  1. Download and unzip your skill package.
  2. Open the `openclaw/` directory inside the zip.
  3. Copy the entire skill folder (containing manifest.yaml and playbook.md) into your OpenClaw skills directory.
  4. The skill appears automatically in OpenClaw's available skills list.
  5. OpenClaw loads the manifest for metadata and the playbook for execution instructions.
  6. Trigger the skill via conversation or automation — OpenClaw handles tool routing and execution.

Tips

  • manifest.yaml defines metadata, category, tags, and required tools.
  • playbook.md contains the full skill instructions that OpenClaw follows.
  • Skills can declare tool dependencies (web_search, exec, etc.) in the manifest.