oracle

Bundle prompts and repository files into one-shot requests for external AI models.

Updated Feb 16, 2026
One-click install
npx skills add https://github.com/travis-burmaster/agentbox --skill oracle-travis-burmaster
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: oracle
Source: https://github.com/travis-burmaster/agentbox/tree/main/agentfork/skills/oracle
Command: npx skills add https://github.com/travis-burmaster/agentbox --skill oracle-travis-burmaster

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @steipete/oracle.

What problem does it solve? Getting a second opinion from a powerful external model requires manually copying code, context, and questions into a chat window. This Skill teaches best practices for the oracle CLI, which bundles your prompt plus selected files into a single one-shot request so another model can answer with real repository context. ## Core Features & Use Cases - Prompt + File Bundling: Attach files, directories, and globs with include/exclude patterns, honoring .gitignore and size limits. - Engine Selection: Use browser automation for GPT-5.2 Pro long-think sessions or the API engine for Claude, Grok, and multi-model runs. - Session Management: Reattach to detached or long-running stored sessions instead of re-running expensive queries. - Use Case: You hit a stubborn bug in a large codebase. Preview the token cost with a dry run, then send the relevant source files plus a detailed prompt to GPT-5.2 Pro via browser mode, and reattach to the session an hour later to read the advisory analysis. ## Quick Start Ask the agent to preview and run an oracle query with a dry run over your source directory, then launch a browser-engine session with GPT-5.2 Pro using a focused file set.

Frequently Asked Questions about oracle

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I send code files to ChatGPT with the oracle CLI?

Use oracle with --engine browser and attach files via --file flags, for example: oracle --engine browser --model gpt-5.2-pro -p "your task" --file "src/**". The CLI bundles your prompt and files into one request and stores the session for later reattachment.

How do I preview token usage before running an oracle query?

Run oracle with --dry-run summary plus --files-report to see the payload and token spend without sending anything. Add -p "<task>" and your --file patterns to get an accurate estimate before committing to a run.

What is the difference between oracle API and browser engines?

The browser engine drives ChatGPT or Gemini through browser automation and suits long GPT-5.2 Pro sessions. The API engine is required for Claude, Grok, Codex, or multi-model runs, and is auto-selected when OPENAI_API_KEY is set.

What should I do when an oracle browser session times out?

Do not re-run the query, since browser runs with GPT-5.2 Pro commonly take 10 minutes to an hour. List stored sessions with oracle status --hours 72 and reattach using oracle session <id> --render.

Which files does oracle exclude when attaching directories?

Oracle skips node_modules, dist, coverage, .git, .turbo, .next, build, and tmp by default, honors .gitignore, filters dotfiles unless explicitly matched, rejects files over 1 MB, and does not follow symlinks.