parallel-dimensional-audit-pattern-for-skill-corpus-quality
Auditing a large skill corpus (71 skills, 11,723 lines) requires parallel agents across orthogonal dimensions rather than sequential sweep: cross-reference integrity, description trigger quality, progressive disclosure / size discipline, information density / anti-slop, and adversarial critic. Each agent covers a non-overlapping quality axis. Single-agent sequential audit misses category-specific patterns.