rover-001/checkmate: Goal-driven mode for AI coding agents. Reach the user's win condition in the fewest moves.
Checkmate is a rule set for AI coding agents (Claude Code, Cursor, OpenCode, Codex) that pushes them to identify the win condition silently, take the shortest path, skip preambles, menus and unrequested files, and verify before saying "done." It's enforced through a SessionStart hook rather than a passive skill file, since agents rarely self-invoke skills, and comes with lite/full/ultra modes plus review, help and gain companion skills. Its own 40-task benchmark claims a higher pass rate than baseline (98.3% vs 86.7%) at roughly half the output tokens and steps, with medians computed only over passing runs and a shared-task comparison to limit survivorship bias. It positions itself as a broader successor to "ponytail," optimizing moves to goal rather than lines of code.