Skip to content

GitHub requests and tokens

This page states which GitHub requests the CLI makes and how it authenticates them. Before you start shows how to set a token for adoption.

Public source acquisition uses a bounded set of GitHub REST requests to resolve the repository, immutable tag, commit and tree, then obtains all source objects through Git’s anonymous smart protocol. The number of REST requests does not grow with the source’s file count. The CLI keeps each verified source outside the project, in $XDG_CACHE_HOME/repo-standards/ or ~/.cache/repo-standards/, and reuses it only once its bytes match their Git identities. A later command or answer round then makes two REST requests, plus one per annotated tag object, resolving the repository and its tag afresh. A complete guided adoption with a lightweight tag makes at most 16, plus one per additional page of releases. Annotated tags add one request per object on each observation, including start’s second observation after it acquires the runtime. Deleting the cache only makes the next command acquire the source again. Every GitHub REST request, including inspection, start, source search and the outdated releases lookup, uses GH_TOKEN, or otherwise GITHUB_TOKEN, as bearer authorization when present. Values are trimmed; empty or whitespace-only values count as absent. A token uses the authenticated API quota instead of the shared anonymous quota and is never sent on Git object downloads or recorded in output, retained inputs or caches. Private sources remain unsupported. An exhausted quota fails inspection, start and search with QUOTA_EXHAUSTED, naming the retry time or delay when available and both token variables, preferring retry-after over x-ratelimit-reset when both arrive. Retry after the reported reset rather than repeatedly restarting a failed command. A rejected token (HTTP 401) is SOURCE_UNAVAILABLE, without anonymous retry. outdated reports either failure as unknown and still exits 0. Credentials are not required for supported public sources.