A Claude AI skill that reviews and fixes clause numbering and stale cross-references in legal contracts and agreements.
94
Pending
Does it follow best practices?
Impact
94%
0.98xAverage score across 10 eval scenarios
Pending
The risk profile of this skill
Design a solution that verifies not only that each cross-reference points to an existing clause, but also that the referenced clause is semantically correct — i.e., the clause actually describes what the surrounding text implies it should. The solution must read both the text surrounding each reference and the content of the referenced clause to assess whether the match is genuine.
The solution should process each cross-reference extracted from the document and perform a two-part semantic check:
The solution then compares the stated intent (from surrounding context) against the actual content of the referenced clause. If the content matches the intent, the reference is marked correct. If the content clearly does not match, the reference is marked incorrect and a suggestion for the likely correct clause may be offered. If the solution cannot confidently determine whether the content matches, the reference must be marked uncertain and flagged for manual review — no change should be made in uncertain cases.
Accepts the full cross-reference list, the clause map, and the full document text. Returns, for each reference, a verdict object containing:
verdict: one of correct, incorrect, or uncertainexplanation: a brief explanation of why the verdict was assignedAI skill that reviews and fixes clause numbering issues and stale cross-references in legal contracts.
docs
evals
scenario-1
scenario-2
scenario-3
scenario-4
scenario-5
scenario-6
scenario-7
scenario-8
scenario-9
scenario-10