Skip to content

GitHub Repository Orientation

Use this workflow for repository onboarding, architecture orientation, locating ownership or conventions, and estimating change impact. Do not use it for a formal pull-request review or to modify repository state.

Identify the repository, the user’s goal, the relevant product or component, and the desired depth. If the repository is ambiguous, resolve the owner and name before reading files.

The upstream names below are capability labels. Select the currently advertised MCP tool whose description identifies that upstream capability.

  1. Use get_file_contents without a ref on the root so the initial view follows the repository default. Read the primary README, contributor guidance, build manifests, and local agent instructions.
  2. Use bounded list_branches and list_commits calls when branch shape or current activity matters to the user’s goal.
  3. Use list_issues and list_pull_requests with bounded filters to identify related active work, not to summarize the entire backlog.
  4. Use search_code for concrete concepts, entrypoints, public contracts, and tests. Follow references with get_file_contents.
  5. Trace one end-to-end path relevant to the user’s question and distinguish direct evidence from ownership or architecture inference.

Read repository-specific instructions before drawing conclusions. Do not assume the newest branch is the default or that directory names prove runtime ownership. If search is incomplete or files are truncated, state what was not inspected.

Provide a compact repository map, key entrypoints, important conventions, active related work, likely change surface, and the next files or tests to inspect. Link to concrete files, issues, pull requests, and commits when the tool returns usable links.

This workflow is read-only. Do not create branches, push files, comment, merge, or update issues or pull requests. Treat private repository content as private and avoid reproducing secrets or sensitive source unnecessarily.