oracle

Get a second opinion by bundling a prompt + a curated file set, then asking another powerful LLM for debugging, refactor advice, design checks, or cross-validation.

⭐3stars🔀0forks📁View on GitHub🕐Updated Jan 9, 2026⬇️4downloads

When & Why to Use This Skill

Oracle is a specialized cross-agent skill designed to provide a 'second brain' perspective by bundling specific code context and prompts for validation by alternative high-performance LLMs. It enhances development workflows by enabling multi-model cross-validation, allowing developers to leverage the unique reasoning strengths of different agents (like Claude and Codex) for debugging, refactoring, and architectural checks.

Use Cases

  • Logic Cross-Validation: Verify complex algorithms or business logic by sending the context to a secondary high-reasoning model to identify potential edge cases or hallucinations.
  • Deep Debugging: Bundle a curated set of source files and error logs to get a fresh diagnostic opinion when the primary agent reaches a bottleneck.
  • Refactor Assessment: Submit proposed code changes to a different model to receive objective feedback on maintainability, performance, and adherence to design patterns.
  • Architectural Design Checks: Use the 'Oracle' to perform sanity checks on system designs or API contracts before implementation, ensuring the proposed structure is robust.
  • Quality Assurance: Perform a final 'second opinion' pass on critical code paths to ensure high-security standards and code quality before merging.
nameoracle
descriptionGet a second opinion by bundling a prompt + a curated file set, then asking another powerful LLM for debugging, refactor advice, design checks, or cross-validation.

Oracle (cross-agent)

Use this skill when you want a fast “second brain” pass from the other agent:

  • In Codex: bundle context, then ask Claude CLI (Opus 4.5) for review.
  • In Claude: bundle context, then ask Codex CLI (GPT‑5.2, xhigh reasoning) for review.

This skill uses a tiny local bundler script.

Workflow

  1. Pick the smallest file set that contains the truth (avoid secrets by default).
  2. Verify the selected files / bundle look right.
  3. Run the Oracle script for your current agent.

Commands

  • When running in Codex:

    • Preview selection: "$HOME/.agents/skills/oracle/scripts/oracle-bundle" --dry-run -p "<task>" --file "src/**" --file "!**/*.test.*"
    • Preview bundle: "$HOME/.agents/skills/oracle/scripts/oracle-bundle" -p "<task>" --file "src/**" --file "!**/*.test.*"
    • Run: "$HOME/.agents/skills/oracle/scripts/oracle-to-claude" -p "<task>" --file "src/**" --file "!**/*.test.*"
  • When running in Claude or any other agent:

    • Preview selection: "$HOME/.agents/skills/oracle/scripts/oracle-bundle" --dry-run -p "<task>" --file "src/**" --file "!**/*.test.*"
    • Preview bundle: "$HOME/.agents/skills/oracle/scripts/oracle-bundle" -p "<task>" --file "src/**" --file "!**/*.test.*"
    • Run: "$HOME/.agents/skills/oracle/scripts/oracle-to-codex" -p "<task>" --file "src/**" --file "!**/*.test.*"

Tips

  • Prefer a minimal file set over “whole repo”.
  • If you need diffs reviewed, paste the diff into the prompt or attach the diff file via --file.
  • Make the prompt completely standalone: include error text, constraints, and the desired output format (plan vs patch vs pros/cons).
  • Never include secrets (.env, tokens, key files).
  • Oracle can be slow while it reasons. Allow it several minutes to process.