Curbside Consult

Use this when a clinical answer needs arithmetic: CrCl (Cockcroft-Gault), IBW/AdjBW/BMI/BSA, vancomycin AUC, aminoglycoside interval, opioid MME, steroid equivalents, corrected Ca/Na, or anion gap (tested python3 tool, adults).

Use this when a clinical question needs live, citable evidence: PubMed or Europe PMC literature, FDA drug labels, trial registries, LactMed, or LiverTox. It covers how to query each free public API and how to report lookups that fail or come back empty.

Use this when asked to verify a reference list or the citations in an article, slide deck, policy, or draft. It checks each reference against Crossref and PubMed for existence, matching metadata, and retraction, and gives each one an honest verdict.

Use this when a question asks what the current FDA label says about dosing in renal impairment (a CrCl value), hemodialysis, peritoneal dialysis, CRRT, or hepatic impairment. It is a procedure for pulling the live DailyMed/openFDA label and quoting the dosing text verbatim with set id, version, and date.

Use this when reviewing any clinical document (job aid, order set, protocol, policy, slides, handout, note template) for Joint Commission do-not-use items, ISMP error-prone abbreviations and dose designations, and tall-man lettering. It is a checklist procedure; findings are candidates to confirm in context.

Use this when the owner provides an institutional document (renal dosing protocol, formulary or restriction list, antibiogram, order-set dosing table) that should outrank national sources, and when answering from one already loaded. It is a procedure: cell-by-cell verified CSV plus a notes file under /workspace/curbside/local/, then answers local-first with a label comparison.

Use this when interpreting culture and susceptibility results: intrinsic (expected) resistance, AmpC-inducible organisms, whether a blood-culture or other isolate is a likely contaminant or a pathogen, and how to handle MIC breakpoints. It gives compact rules and says what must be looked up live.

Use this when the owner asks to test Curbside Consult, after installing or updating its skills, or when behavior seems off. It runs the calculator's selftest and walks through hard prompts with the expected safe behavior, then reports PASS/FAIL.

Use this in a Curbside Consult bot's first conversation with a new owner: a short setup (role, region, defaults, answer length) saved to memory, plus installing and self-testing the calculator. Every question can be skipped.