battery

What problem does it solve? Single-reviewer audits of a game's spec, design, or build miss whole classes of defects because one perspective cannot cover fun, economy math, UI polish, save integrity, and canon drift at once. This Skill orchestrates independent fresh-eyes agents that each attack the subject through a different lens, then a convergence critic de-duplicates and scores the findings. ## Core Features & Use Cases - Three audit sizes: a full battery (6-13 lenses for milestone or pre-ship gates), a mini battery (~3 lenses to validate one fix or scenario), and a diff re-audit (~3 lenses over just the delta after a locked-ADR execution or large refactor). - Structured lens pipeline: lens agents read the subject cold and return structured findings (severity, evidence, proposed fix, new-vs-carryover), with adversarial refutation checks for large claims and a convergence critic that assembles the scorecard. - Full archiving and routing: raw verdicts are snapshotted, synthesis reports land in project/audit/reports/, a lens registry enforces novelty, and every actionable finding becomes an ADR or human-in-the-loop ledger row. - Use Case: Before calling a large refactor done, run a diff re-audit so three independent lenses verify the change matches the ADR, introduced no new failure modes, and did not silently alter an approved design pick. ## Quick Start Run a mini battery on the latest combat rebalance to validate the fix before I call it done.