puzzle-generator

Word grid puzzle generation here uses a generate, review, and fix loop with subagents to enforce at least 3 distinct relation types and 6+ A-tier edges. It outputs a 3x3 word grid with 12 edges and per-edge clues.

To generate a 3x3 word grid with 12 edges, provide seed anchors to start the pipeline. The Skill then maps word connections, applies edge taxonomy classification, and returns a TypeScript puzzle object with emoji annotations.

Yes, you can ensure sufficient relation type diversity using the built-in generate, review, and fix loop. This workflow includes blind audits and reviewer checks to guarantee at least 3 distinct relation types and 6+ A-tier edges.

Yes, the puzzle generator outputs a TypeScript puzzle object that references the repository's puzzle model and metadata structure. It includes generation metadata fields, accompanying edges, and emoji annotations for each word.

Automated emoji puzzle generation is limited to 3x3 grids with 9 words and 12 edges. The pipeline relies on strict edge taxonomy classification and seed-anchor checks to maintain board diversity and solvability.