Problem

Two architecture issues identified in docs/ARCHITECTURE-IMPROVEMENTS.md:

1. tests/run-all.js uses a hardcoded list of test files. New tests are never run unless someone manually updates the list.
2. Agent/skill/command counts drift between README.md, AGENTS.md, and actual filesystem contents.

Scope

1. Replace hardcoded test list in tests/run-all.js with glob-based discovery (**/*.test.js under tests/)
2. Add a catalog sync script (e.g. scripts/ci/catalog.js) that scans agents/*.md, commands/*.md, skills/*/SKILL.md and verifies counts match documentation

Acceptance Criteria

• New test files are automatically discovered and run
• CI validates documentation counts match filesystem reality
• README/AGENTS.md counts are accurate