Skip to content
Get packs

Oracle Second-Model Review

Bundle your prompt and repo files and get a second model's advisory review via Oracle CLI

Source Download skill
Repository
steipete/agent-scripts
Stars
7,433
Licence
MIT
Files
1 file

What it does

This skill teaches the agent to use the Oracle command-line tool, which packages your question plus selected project files into one request and sends it to another AI model (such as GPT-5.5 Pro in ChatGPT through browser automation, or a model via API) for a second opinion on debugging, refactoring or design. The agent learns how to choose a tight file set, preview what will be sent, run long browser sessions and reattach to them, and handle API keys and costs carefully. The skill notes that the outputs are advisory and should be checked against the codebase and tests.

Great for

  • Developers who want a second model to review a tricky bug, refactor or design decision using real repository context
  • People already paying for ChatGPT Pro who are comfortable with a slow, human-in-the-loop browser workflow
  • Teams that want to compare answers from several models in one advisory panel

Skip it if

  • Users who need fast answers or cannot wait up to an hour for browser-based runs
  • Projects containing secrets or sensitive code that should not leave the machine, since selected files are sent to third-party model providers
  • Anyone expecting the second model's output to be automatically correct, as the skill itself says outputs must be verified

Try asking

  • “Use Oracle to review src/ for the root cause of this failing login bug, previewing the file set first”
  • “Do a dry run showing token counts for the payments module before sending it to a second model”
  • “Ask a second model to suggest a refactor of the auth code using GPT-5.5 Pro in the browser”
  • “Run an advisory panel across GPT, Claude and Gemini via the API with partial results saved to a file”
Step-by-step guide: how to use Oracle Second-Model ReviewInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    oracle.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip oracle.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip oracle.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Always run --dry-run with --files-report first to catch large or unwanted files before spending tokens or time
  • Exclude tests, snapshots and secrets with ! patterns, and avoid sending .env or credential files
  • If a browser run times out, reattach to the stored session with oracle session <id> rather than re-running it

What's inside

  • SKILL.md

Redistributed unmodified from steipete/agent-scripts under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.