resolve-staging-proposals

What problem does it solve?

This Skill solves the problem of unlinked staging records produced by Dutch government data parsers (parse-abd-nieuws, parse-staatscourant, parse-organogram) that cannot be automatically merged into the canonical Polder dataset, as they lack references to existing organization, post, and person entries.

Core Features & Use Cases

• Automated Entity Matching: Matches staging proposal records to existing canonical entries in data/organisaties/, data/posten/, and data/personen/ using exact and fuzzy name matching, parent relationship validation for organizational subdivisions, and person name parsing with initials and birth year support.
• Confidence & Merge Logic: Assigns per-field confidence scores for organization, post, and person matches, enforces the two-source rule for state gazette and ABD news sources, and generates merge recommendations (auto-merge, needs-review, skip) to streamline the daily government data update workflow.
• Use Case: When processing a new staging file with a proposed appointment for an afdelingshoofd, the Skill automatically links the proposed organization to the existing Directie Wonen record, flags missing posts for creation, and marks low-confidence or ambiguous matches for human review.

Quick Start

Use the resolve-staging-proposals skill to match a staging file from the data/_staging directory to existing canonical government records and receive enriched output with resolved IDs, confidence scores, and merge recommendations.