hoo-citation-trackerlisted

# AI citation tracking 1. Read profile: keywords.targets (the query set), competitors, site url, and the optional `citations: {engines: [...]}` key (see Engine set below). 2. For each query (cap 20 per run; rotate through the set across runs), ask each engine in the engine set that is actually reachable from this session. Sources in order of preference: an authorized AI-search connector; the search-data adapter's AI-visibility read, when its tools are present in this session (the search-data slot, ADR-0009 in the repo; the known adapter is OpenSEO - see `$CLAUDE_PLUGIN_ROOT/docs/connectors.md` - licensed data, and every line carrying it names the adapter as its source); WebSearch with engine-targeted queries; plain WebSearch otherwise. Never scrape engines through automation that violates their ToS. 3. Record per query: engine | our domain mentioned? | cited (linked)? | position (only when cited - the ordinal below) | sentiment (only when mentioned - the label plus its evidence quote, below) | competitors mentioned | answer summary (<=30 words). 4. Metrics: mention rate, citation rate, share of voice vs competitors (mentions of us / mentions of anyone tracked). 5. Persist to runs/YYYYMMDD-citations/ + append one signal line with the headline movement vs the previous run (diff the last runs/ folder). 6. New citation appearing or disappearing on a money query -> P1 signal. 7. When `runs/<date>-ai-baseline/REPORT.md` exists in the