Oracle Code Reviewer
Bundles repository context into high-fidelity prompts for second-model code reviews and architectural cross-validation.
Oracle is a specialized CLI-driven skill that allows Claude to package specific project files and directories into a single 'one-shot' request for review by secondary AI models. By utilizing either API connections or browser automation, it facilitates deep-dive debugging, complex refactoring, and design checks that benefit from a second opinion. It features robust session management, token budget monitoring, and smart context filtering to ensure that external models receive the most relevant information without exceeding limits or compromising security.
Características Principales
01Session Persistence: Stores and reattaches to long-running review sessions to prevent work loss.
02Token Budgeting: Provides dry-runs and file reports to estimate token usage before execution.
03Smart Filtering: Respects .gitignore and automatically excludes build artifacts and dependency folders.
040 GitHub stars
05Multi-Engine Execution: Seamlessly switches between API-based reviews and browser automation.
06Context Bundling: Packages prompts with files, directories, and globs into a single request.
Casos de Uso
01Cross-validating major architectural refactors with secondary high-reasoning models.
02Deep debugging of repository-wide issues by providing full context to a 'human-in-the-loop' browser session.
03Generating comprehensive patch plans and test cases by feeding context-rich prompts to specialized LLMs.