OpenScite

Analyze papers citing one user-provided scholarly PDF. The uploaded PDF is the target, not a source of references to follow. Preserve user filenames and write the report in the user's language while keeping titles and quoted passages in their source language. Use keyless OpenAlex.

Use the resumable runner

Resolve this skill's directory as <skill-dir>. Run commands from the user's workspace so default artifacts are created there. Do not generate a substitute pipeline script.

python "<skill-dir>/scripts/openscite.py" prepare "<target.pdf>" --language <zh-TW|en> [--n N] [--run-dir <dir>]

Make prepare the first PDF operation. It runs the latest local Anydoc release, caches the output, matches arbitrary inbox filenames by DOI/title, and falls back only when its artifacts record an extraction failure. Add --require-page-aware only when exact page locators matter; this intentionally uses the page-aware fallback. Use --rule-triage only for diagnostics.

Reuse the same target, run directory, N, mode, and language on every retry. Follow the returned status: