oracle

To bundle repository files with a prompt, combine your instructions with selected source files and config into a single context package. This creates a high-context request for accurate external-model analysis, reducing guesswork during bug triage or architecture reviews.

Prompt bundling for repository investigation is the process of packaging instructions alongside selected source files into one context package. It ensures an external model has the exact file context needed to analyze bugs, review architecture, or generate patch plans accurately.

To reattach to a stored browser session, use the session management feature to reconnect to stored runs instead of restarting. This handles browser-based requests that detach or take a long time, preserving long-form reasoning progress.

Yes, prompt bundling requires explicit file selection to ensure safe handling of attachments without exposing secrets. You must manually choose the relevant source files and config, and preview tokens before sending the context package to the model.

Yes, you can choose between a browser engine or API workflows for file-aware reasoning. Engine selection depends on whether you want long-form reasoning, automation, or model flexibility when analyzing repository files.

Limitations of bundling repository context include the requirement for explicit file selection and token previewing to avoid exceeding limits. Safe handling of attachments demands manual secret redaction, and long browser-based runs may detach, requiring session reattachment.