Clause Map Building

Capabilities { .capabilities }

Design a solution that builds a complete clause map from a legal contract document. The clause map must enumerate every top-level clause and every sub-clause, recording both the clause number and a short heading/description. This map will be used as ground truth for all subsequent analysis steps.

• Given a document with top-level clauses (e.g., "1. Definitions", "2. Term", "10. Holiday Entitlement"), the clause map includes every top-level clause with its number and heading @test
• Given a document with sub-clauses (e.g., "10.1 Annual leave entitlement", "10.4 Carry-over of unused days"), the clause map includes every sub-clause with its number and short description @test
• Given a .docx document, the solution parses the full text programmatically (not by screenshot) to produce the clause map @test
• The produced clause map is maintained as the ground truth reference throughout the session and not discarded after initial analysis @test

Implementation { .implementation }

The solution should open the target document (Google Doc or .docx) and extract a structured list of every clause it contains. For each clause, it records the clause number (e.g., 10, 10.1, 10.4) and the associated heading or short description. For Word documents, extraction must be performed programmatically by iterating over the document's paragraphs using python-docx — reading screenshots is not sufficient. Once built, the clause map must be retained in memory and referenced in all subsequent steps; it must not be rebuilt from scratch or discarded mid-session.

API { .api }

Accepts a document (Google Doc or .docx) and outputs a structured clause map. Each entry in the map contains:

• clause_number: the numeric identifier (e.g., "10.1")
• heading: a short heading or description for the clause

Dependencies { .dependencies }

contract-numbering-fixer 0.1.0 { .dependency }

AI skill that reviews and fixes clause numbering issues and stale cross-references in legal contracts.