A specialized CLI tool that bundles prompts and source files into a single context-rich request for advanced AI model reviews and debugging.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install oracle
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 oracle 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 Oracle AI Context Bundler?
Oracle is a sophisticated command-line utility designed to bridge the gap between your local codebase and high-reasoning AI models. By leveraging Openclaw Skills, Oracle packages your specific task instructions alongside relevant source files into a "one-shot" bundle. This ensures that the AI model—whether accessed via API or browser automation—has the full repository context needed to provide accurate debugging, refactoring, or architectural advice.
Rather than relying on the limited context windows of standard chat interfaces, Oracle treats your codebase as a first-class citizen. It allows developers to provide the "truth" of their project to models like GPT-5.2 Pro or Gemini, facilitating deeper insights that typical AI agents might miss. This tool is essential for anyone looking to integrate high-fidelity Openclaw Skills into their daily development cycle.
Oracle AI Context Bundler Use Cases
- Deep debugging of complex logic across multiple files using advanced reasoning models.
- Performing comprehensive code refactors where project-wide context is essential for maintainability.
- Conducting design checks and cross-validation against existing codebase conventions and styles.
- Generating one-shot review requests for models like GPT-5.2 Pro through automated browser sessions.
- Creating portable Markdown bundles of code context for manual sharing or archival.
How Oracle AI Context Bundler Works
- Select a specific set of files or directories using flexible glob patterns and automated exclusion rules.
- Perform a dry run to verify the token count and file selection before sending any data.
- Execute the context bundle through either the browser engine (for ChatGPT/Gemini) or the API engine (for Claude/Grok).
- Review the AI-generated advisory output and verify the suggestions against your local tests and codebase.
- Reattach to existing sessions if a long-running browser process detaches or times out to avoid losing work.
Oracle AI Context Bundler Setup
Oracle can be run directly using npx without a permanent global installation.
# Display help and available command-line options
npx -y @steipete/oracle --help
# Run a basic preview to see what will be sent
npx -y @steipete/oracle --dry-run summary -p "Explain this logic" --file "src/**"
# Execute a browser-based run for GPT-5.2 Pro
npx -y @steipete/oracle --engine browser --model gpt-5.2-pro -p "Refactor the auth logic" --file "src/auth/**"
Oracle AI Context Bundler Data Schema & Taxonomy
Oracle manages sessions and context through a structured local file system and metadata taxonomy:
| Component | Description |
|---|---|
| Session Storage | Files are stored under ~/.oracle/sessions by default. |
| File Expansion | Supports globs, literal files, and directories while honoring .gitignore. |
| Default Ignores | Automatically skips node_modules, .git, dist, and build unless specified. |
| Token Management | Tracks input volume with a target limit of approximately 196k tokens. |
| Session IDs | Uses unique IDs or user-defined slugs for easy session retrieval. |
Oracle AI Context Bundler Advanced Features
- Remote browser host support, allowing automation to run on a separate signed-in machine.
- Detailed token and cost sanity checks via the
--files-reportand--dry-run jsonflags. - Advanced attachment handling with
--browser-attachmentsto manage inline pasting vs. file uploads. - Persistent session management with
oracle statusand the ability to reattach to disconnected runs. - Hidden verbose modes and manual rendering options for power users to customize the prompt injection.
SKILL.md
Loading
Related Openclaw Skills
OpenHue
A powerful CLI integration for controlling Philips Hue smart lighting systems directly through your AI agent.
OpenAI Whisper CLI
A local speech-to-text tool using the OpenAI Whisper CLI for private, high-accuracy audio transcription without the need for API keys.
OpenAI Whisper API Transcription
A high-performance CLI skill for transcribing audio files using OpenAI's Whisper model via the Openclaw Skills framework.
OpenAI Image Gen
Batch-generate AI images using OpenAI with automated prompt sampling and a built-in HTML gallery.
OrderCLI Foodora & Deliveroo Assistant
A command-line interface for tracking active food deliveries, viewing order history, and performing reorders on Foodora and Deliveroo.
Peekaboo
A comprehensive macOS UI automation CLI designed to capture screens, inspect elements, and drive user inputs for AI agents.
Featured*
