# Verification

## What should be true

- The skill directory exists.
- The skill has a clear workflow description.
- The runtime layer can load role JSON files from `/root/projects/meeting/agents/`.
- The role suggestion flow exists.
- The background ingestion guidance exists.
- The message style rule says each role turn is a separate message.

## Quick manual checks

1. Open `SKILL.md` and confirm it references:
   - final spec
   - quick start
   - ingestion guidance
   - runtime scripts

2. Open `references/final-spec.md` and confirm it includes:
   - AI participant suggestion
   - human confirmation
   - standalone speaker turns
   - large text handling

3. Open `references/ingestion.md` and confirm it includes:
   - paste text
   - file/document source
   - sectioned background
   - AI summarization

4. Open `scripts/runtime.py` and confirm it includes:
   - role loading
   - meeting plan/state
   - PM decision model
   - single-role analysis

5. Open `scripts/handlers.py` and confirm it includes:
   - AI role suggestion
   - role parsing
   - PM routing
   - role execution