monitorlisted

# Monitor ## Goal Turn one-off measurement into a standing watch. Setup writes a `citable/monitor.json` config and schedules a weekly run; each run answers two questions — *did rankings move?* and *did AI answers change who they cite?* — against last week, not against nothing. Setup spends nothing; only scheduled runs spend, and never more than the budget in the config. ## Two modes - **Setup** (a person asked for a monitor): gather inputs, quote the recurring cost, write the config, install the schedule. No paid calls. - **Run** (`monitor run`, invoked by the scheduler): read `citable/monitor.json`, execute the run contract below, append history, report the diff. No questions asked, nothing outside the config. ## Setup Gather, in one message, only what is missing: 1. The domain, and the keyword list to track — or take it from `rank-tracking`'s history or a `keyword-brief` result if one exists in the repo. 2. Buyer questions **the user wrote** (0–10) and which engines, for the AI-citation leg. No questions → the leg is off, not invented. 3. Cadence — default weekly. Daily re-buys engine noise; say so if asked. Then quote the standing cost before anything is scheduled: ``` per run = keywords × $0.012 + questions × engines × $0.05 ``` State it as $/run and $/month, check the wallet covers at least 4 runs (`citable_prices`, free), and get one yes — it covers every future run **at this config**. Any change to keywords, questions, engines, or cadence is a new quote.