Usage

Use synthetic question ID 123456 below in place of your own question ID. Exactly one content source is required:

hackerrank questions upload 123456 --directory ./starter --dry-run
hackerrank questions upload 123456 --directory ./starter --exclude '*.zip'
hackerrank questions upload 123456 --file ./starter.py

Real uploads read HACKERRANK_API_TOKEN from the environment. Get a key from the HackerRank for Work API tokens page. Missing, empty, and whitespace-only keys fail clearly. Dry runs need no key and make no network requests.

Exactly one of --directory or --file is required. Both replace the entire project archive. --file creates a one-file project using the source’s basename; it does not update a language-specific code stub or upload a prebuilt ZIP as the project archive.

All file bytes are preserved exactly, including binary files, CRLF, blank lines, trailing whitespace, UTF-8 BOMs, and documentation markers. Empty single files are valid. Neither paths nor IDs are inferred from repository configuration. Inputs are fully read, staged, and archived before credentials are read or a mutation is issued. POSIX executable bits are preserved. Errors exit nonzero. API response bodies and credentials are never printed. The default retry count is zero, matching the released SDK. Set --retries 3 to delegate retries and backoff to the SDK for this repeatable upload. The CLI adds no separate retry loop. Each attempt uses the same prepared archive bytes.

Directory selection

• Both Git and non-Git directories work. Paths are relative to the current working directory. A Git worktree’s .git file also identifies its root.

• In a Git repository, rules are inherited from the nearest Git root down to the upload directory. Outside Git, rules start at the upload directory; unrelated parent .gitignore files are not used.

• Each traversed directory adds its own .gitignore rules. Later matching rules override earlier ones, and deeper files override ancestor rules. Patterns are relative to their owning directory. Directory-only rules and negations follow Gitignore semantics. Excluded directories are pruned, so a child cannot re-include itself unless its parent is re-included first. This also applies to ancestors of an explicitly selected nested upload root.

• Ignore rules apply to tracked files too. Global Git excludes and .git/info/exclude are not read. Hidden files, including .gitignore, are uploaded unless excluded.

• .git files and directories are always excluded at every depth. Explicit source paths inside Git metadata are rejected, including an upload rooted at a .git directory. Repeatable --exclude PATTERN options form a final Gitignore rule layer, relative to the upload root. This layer can exclude files re-included by .gitignore; its own later negations can undo its earlier patterns. It cannot re-include a file excluded by .gitignore or Git metadata.

• ZIP files are included by default. Use --exclude '*.zip' to omit them. Add other exclusions explicitly, such as --exclude node_modules/.

• Selected symlinks (including directory, broken, and external links) and symlinks in the source path are rejected rather than dereferenced. Ignored links are skipped. Symlinked .gitignore files are rejected when their rules would be read. Special files are rejected. Empty selections fail. Selected files are copied to a temporary directory without dereferencing links. The CLI packages these prepared files with Python’s standard ZIP library. The released SDK handles multipart serialization, authentication, network requests, response decoding, and retries.

Development and distribution

Development checks, release setup, and build instructions for Docker, Nix, and standalone binaries are in docs/source/development.rst. The generated CLI reference is in docs/source/cli.rst. Prebuilt binaries are available from GitHub Releases.

See docs/source/migration.rst for migrating an existing shell uploader and keeping repository-specific snippet preparation outside this CLI.