oracle

To bundle repository files with a prompt for better AI answers, use the oracle CLI to package your task and selected files into a one-shot request. Specify file globs like "src/**" to include relevant code modules so the model responds using actual project context.

Repo-grounded code analysis prevents weak, guessy AI answers by attaching truth-relevant files directly to your prompt. You need it when asking a model to debug, refactor, or explain a specific subsystem where actual project context is required for an accurate patch plan.

Yes, you can use browser automation for long-running repository coding tasks by selecting the browser engine mode. This allows interactive execution and enables session persistence so you can reattach to active browser sessions if a task takes extended time to complete.

Control payload size and exclude files when sending repository context by applying file glob include and exclude patterns during prompt bundling. You can also perform payload dry-runs to preview the request size and estimated costs before executing the targeted model run.

A key limitation of attaching codebase files to AI prompts is the risk of exposing sensitive data. You must apply safety practices to avoid attaching secrets, and carefully manage file glob patterns to prevent accidentally bundling configuration files or unrelated repository context into the one-shot request.