Agent Sort

This does one thing well: it looks at your actual repo and decides which ECC components you need loaded daily versus tucked away in a searchable library. Instead of guessing or manually curating, it greps for evidence like file extensions, package managers, framework configs, and build scripts, then classifies skills, rules, hooks, and commands into DAILY or LIBRARY buckets. The parallel review passes are clean if you have subagents available. Use it when full ECC installs are too noisy for a focused project, or when a repo has drifted into the wrong language ruleset. The evidence table format keeps decisions auditable instead of vibes based.