The combined restaurant audit: runs all THREE sibling skills on one outlet list and renders a single interactive page scoring each place on (1) food authenticity & taste (from food-review-review), (2) food quality - organic & seed-oil-free (from find-ultra-quality-food), and (3) air conditioning (from find-restaurants-with-ac). The page has a sort-by dropdown: the chosen dimension becomes the primary lexicographic sort key, the other two following in the default order taste > food quality > AC. Use when the user wants the all-in-one view of the best places among an explicit list OR a category + location (e.g. "Indian restaurants in Norwich", "all restaurants in Diss"). Token-intensive (it includes the full food-review-review pass).
Find which restaurants/takeaways have air conditioning. For each outlet, answer ONE binary fact - does it have AC? - with a confidence score (how much to trust the answer, from the number, recency and type of sources), by reading the outlet's website verbatim, its Google/ TripAdvisor amenity listing, articles and (especially) reviews, with parallel full-strength agents (one per outlet). Produces a super-simple ranked table: "Has AC" + confidence. Use when the user wants air-conditioned places among an explicit list OR a category + location (e.g. "restaurants in Diss", "cafes in Norwich", "restaurants in Seville").
Audit a set of restaurants/takeaways for two specific health-quality facts about HOW the food is made - (1) are there seed-oil-free options? and (2) are there organic options? - by reading their websites, menus (downloading and OCR'ing PDFs/images), articles and a light pass of reviews with parallel full-strength agents (one per outlet). Each fact gets a yes/no/unknown verdict plus a reliability score (how much to trust it, from the number, recency and type of sources). Produces a ranked, easy-to-scan HTML report. Use when the user wants to find seed-oil-free and/or organic places among an explicit list OR a category + location (e.g. "Indian restaurants in Norwich", "all takeaways in Diss", "restaurants in Paris"). Narrower and cheaper than food-review-review - it extracts two binary facts, it does not cluster every review.
Audit and RANK a set of restaurants by genuine food quality and authenticity by reading EVERY Google and TripAdvisor review (harvested via Apify) and clustering each one with parallel full-strength agents — never sampling, never an ML/embedding shortcut. Filters out non-food gripes (delivery, wait, price) and unsophisticated-palate complaints to surface the credible food signal, then produces a ranked, visual HTML report. Use when the user wants to find the best authentic food outlets among an explicit list of restaurants OR a category + location (e.g. "Indian restaurants in Norwich", "ramen in Brighton"). Token-intensive by design.
Invoked via /setup-claude-token-optimiser to interactively set up a "kick" heartbeat that optimises Claude usage-window tokens for a developer who spreads work across the day. It explains the 5-hour-window trick and when it does/doesn't help, asks the user about their work pattern and sleep preference (with yes/no options), computes an optimal cron schedule, installs the kick script + crontab (non-interactively), and — on macOS — sets up state-aware sleep disabling (offsleep/onsleep + a boot LaunchDaemon), handing the user the sudo commands to run.