What is citecheck?

CiteCheck is a tool for verifying the citations in an academic paper. It can be used in two ways: as an installable skill that works inside AI coding assistants like Claude Code, or as a standalone command-line program. Either way, it takes a LaTeX project or a PDF as input and produces a structured report covering four things: whether each citation is formatted correctly, whether the cited paper can actually be found in public databases, whether each reference is on the same topic as the paper citing it, and whether specific claims made about a reference appear to be accurate. The lookup step queries Crossref, Semantic Scholar, OpenAlex, PubMed, arXiv, dblp, and Google Scholar in sequence to verify that cited papers exist. Thematic and semantic relevance checks work differently depending on the mode: when run as an agent skill, the host AI handles this reasoning using its own language understanding with no separate API key required. In standalone CLI mode, simple heuristics are used by default, with an optional OpenAI key for stronger matching. Output is a Markdown report listing which citations have formatting problems, which could not be verified in any database, how well each reference matches the paper's theme, and which bibliography entries are never cited in the main text. The skill is designed to work across multiple AI assistants. Installation paths are listed for Claude Code, Codex, OpenClaw, Hermes, Gemini CLI, and Cursor. The Python package is also available on PyPI for direct command-line use without any AI assistant involved. The project is in active development and accepts contributions with the requirement that the skill file stays compatible across all supported agents.