legalwork

Create a task for the LegalWork model benchmark from a description. Use when the user wants to "create/add a benchmark task", "make a new eval or test case", "turn this into a benchmark task", or "add this to the benchmark" — i.e. define a legal task with instructions and pass/fail criteria to compare how connected models perform.

Firm-owned PDF reading + actions over the in-app PDF viewer. Use whenever the user wants to work with a PDF: read it (extract its text to quote, summarize, or compare), annotate it (sticky notes, highlight boxes), list or fill its form fields (AcroForm), or stamp a signature (typed name + date, or a PNG signature image). Editing always writes a NEW copy (.annotated.pdf / .filled.pdf / .signed.pdf) — the original PDF is never modified. Self-contained; runs on the firm's own infrastructure; pairs with the in-app PDF viewer.

Firm-owned Word (.docx) reading + editing. Use whenever the user wants to work with a Word document: read/answer questions about a contract ("what's the salary / term / governing law"), check a value against a threshold, redline or comment on a clause, or accept/adopt or reject tracked changes. Reads the FULL document — including clauses laid out in tables — and writes tracked-change redlines + comments as reviewable suggestions anywhere in the document. Self-contained; runs on the firm's own model and infrastructure; pairs with the in-app .docx viewer.

Firm-owned tabular document review — the open-model replacement for Harvey/Legora "review grids". Extract a defined set of fields (columns) across a set of documents (rows) and produce an interactive, source-cited review table. Use whenever the user wants to review/compare/extract across MANY documents at once: due-diligence pulls, NDA/contract abstraction, lease abstraction, "make a table of X across these files", "review these contracts for Y", "build a review grid", or "extract these terms from every document".