Transitive Match Closure

What it does

This skill provides a simple, practical technique to post-process pairwise entity match predictions by enforcing bidirectional links (symmetry) and propagating matches across connected components (transitive closure). It converts independent pair predictions into consistent match groups, improving recall and producing coherent deduplicated clusters for downstream workflows.

When to use it

Use this after a binary pairwise matching model has produced candidate matches but results contain inconsistencies (A→B but not B→A) or fragmented clusters (A→B, B→C but not A→C). Helpful in record linkage, entity deduplication, and any pipeline where consistent grouping of IDs matters. Avoid full closure on extremely noisy predictions unless confidence filtering is applied.

What's included

• Scripts: none bundled in the repo for this skill, but the SKILL.md includes runnable Python snippets demonstrating symmetric_closure and iterative transitive_closure.
• References: minimal; user-provided examples and Kaggle links in the SKILL.md.
• Instructions: parse predictions into an ID→matches map, apply symmetry pass, then apply transitive propagation (or union-find for large datasets). Cap iterations and/or filter by confidence to prevent runaway merges.

Compatible agents

Likely compatible with general-purpose code-capable agents that can run Python snippets (Copilot-like assistants, Claude Code, Cursor).