Universal skeleton: role + context (agency, standards edition, audience) + input data + exact output format + constraints (“cite which sections you’re unsure about”; “list assumptions you made”).

  • Report tightening: “Edit for a Florida traffic study reviewed by [agency]. Keep all numbers/citations verbatim. Active voice, no hedging adjectives, flag any sentence making a claim without a value.”
  • Comment response drafting: “Reviewer comment: [text]. Our analysis did: [facts]. Draft agree/clarify/justify options per Wording Bank and Email Patterns tone; cite [standard] where relevant; I will verify citations.”
  • Excel formula: “Columns: [schema]. Need [logic]. Give formula + a plain-English test I can run on row 2 to verify.”
  • Python: “Task: [transform], input sample: [rows], output: [format]. Include assertions validating input assumptions and print a QC summary.” (→ a code-script note with verification date.)
  • Synchro troubleshooting: “Symptom: [X] in Synchro [ver]. List candidate settings in likelihood order with where each lives and how to test — diagnostic tree, not a single guess.”
  • Crash pattern brainstorm: “Summary table: [paste]. List over-represented patterns worth testing and what cross-tab would confirm each. Screening only — do NOT propose countermeasures yet.”
  • Countermeasure menu: “Pattern: [X] at [context]. List candidate treatments with the FHWA PSC/CMF Clearinghouse entries I should look up — names and where to verify, not effectiveness numbers from memory.”
  • Calming / parking / ULI setup: “Land uses: [program]. Walk me through ULI 3rd-ed factor selection questions I must answer, as a checklist — no numbers, just the decisions and where each factor comes from.”
  • Comparable-site scan: “Criteria: [matrix]. Suggest FL agencies/programs likely to have precedent + search phrasings for agenda packets.”
  • Meeting summary: “Raw notes: [paste]. Output: decisions / actions(who,when) / assumptions stated / open questions. Nothing invented; mark gaps.”
  • Design review second-pass: “Here are my S&PM review comments. Play skeptical senior: what did I likely miss per SPM Plan Review Checklist categories?”
  • CAD/redline help: “MicroStation [ver], FDOTConnect [ver]: how to [task]; give ribbon path AND key-in; list what commonly goes wrong.” Meta-rule: any output touching a deliverable exits through AI Usage Rules #2–5.