A working library of prompts for ChatGPT, Claude and Gemini, written for construction and design work: preconstruction, estimating, architecture, visualisation, engineering, scheduling, site operations, safety, commercial and business development.
Every prompt is a file with the same five sections — when to use it, the prompt itself, what good output looks like, follow-ups that sharpen the answer, and guardrails covering what never to paste and what always to verify. The index below is generated from those files, and CI fails if it drifts, so the library cannot quietly go stale.
- Find the decision, not the topic. The sections are organised around work you are actually doing — pricing a bid, resolving a clash, answering a notice. Start there.
- Fill in the
{{placeholders}}. Every prompt has variables for the facts the model cannot know. Prompts fail far more often from missing context than from missing cleverness. - Read the guardrails before you paste anything. They say what must never go into a consumer tier tool, and what a person still has to check.
- Use the follow-ups. The first answer is a draft. The follow-ups are where the work gets done: asking for the rows, the counter-argument, or the thing you have not considered.
- Keep it as a draft. Everything here produces something a named person reviews before it leaves the company. That is not a formality, it is the control.
Qualifying work, framing the programme, finding the risk before you price it.
| Prompt | Use it when | Tags |
|---|---|---|
| Bid qualification: go or no-go | An invitation to bid lands and you need a defensible recommendation before you spend a week estimating it. | bidding risk preconstruction |
| Constructability risk review of a drawing set | You are pricing or planning a job and want a structured list of buildability questions before the bid goes in. | preconstruction drawings risk |
| Programme and sequencing options | You have a target completion date and a rough scope, and you want the plausible sequencing options on the table before you build a CPM schedule. | preconstruction programme sequencing |
Sanity checks, pricing narratives, and the questions an estimator should be asked.
| Prompt | Use it when | Tags |
|---|---|---|
| Unit rate sanity check | An estimate is nearly finished and you want a second pair of eyes on whether the rates and quantities hang together — before the bid goes in, not after. | estimating pricing review |
| Takeoff review checklist | A takeoff has been produced — by a person, by software, or by one of the tools in our open source construction tools list — and it needs a review that is actually systematic. | takeoff quantities quality |
| Change order pricing narrative | The numbers are agreed internally and now need to be presented so they are hard to argue with — typically for a priced change that a client or a subcontractor will scrutinise. | change-orders commercial pricing |
| Productivity assumption stress test | An estimate depends on production rates you have assumed rather than measured, and you want to know how much of the margin is riding on them. | productivity labour risk |
Briefs, specifications, code checks and design review preparation.
| Prompt | Use it when | Tags |
|---|---|---|
| Design brief from a client conversation | You have come out of a client meeting with scattered notes and need a written brief that the client can confirm or correct before design work starts. | brief client architecture |
| Specification section draft | You need a first draft of a specification section structure so you are editing rather than staring at an empty document — with the understanding that the technical content is yours, not the model's. | specification documentation quality |
| Code and accessibility checklist for a design review | You want a structured set of compliance questions to work through for a design, recognising that the model does not know your local code and must not be treated as knowing it. | code accessibility compliance |
| Design review agenda that finds problems | You are running a design or coordination review and want an agenda that surfaces conflicts rather than walking through the drawings in order. | coordination meetings design |
Image prompts, iteration, and the words that go around a render.
| Prompt | Use it when | Tags |
|---|---|---|
| Photorealistic exterior render prompt ChatGPT, Gemini only |
You need a client-ready exterior visual quickly, before the model is advanced enough for a proper render, or to explore a façade direction. | rendering visualisation marketing |
| Interior render: iteration prompt ChatGPT, Gemini only |
The first interior image is close but wrong — the light, the proportions or the materials — and you want to change one thing at a time rather than regenerate from scratch. | rendering interior iteration |
| Render critique before it goes to the client | You are about to send a visual to a client and want a hard critique first — one that finds the things an architect, a planner or a contractor would spot immediately. | rendering review presentation |
Structural and MEP sense checks, always with an engineer and never instead of one.
| Prompt | Use it when | Tags |
|---|---|---|
| Structural load path sanity check | You are preparing for a coordination or value-engineering discussion and want the obvious load path and stability questions on the table. | structure review engineering |
| MEP coordination and clash triage | You have a clash report or a coordination problem list and need to decide which clashes matter, in what order to resolve them, and who should move. | mep coordination bim |
Narratives, look-aheads and the records a delay claim is built from.
| Prompt | Use it when | Tags |
|---|---|---|
| Schedule narrative for the owner | The monthly schedule update is done and needs the narrative section: what moved, why, what you are doing about it, and what you need — without excuses and without jargon. | schedule reporting communication |
| Three-week lookahead constraint sweep | You are building the next three-week lookahead and want to remove constraints before they stop work, rather than reporting them after the fact. | lookahead field planning |
Turning messy field notes into records somebody can act on.
| Prompt | Use it when | Tags |
|---|---|---|
| Daily report from messy field notes | The day's notes are a mix of texts, voice-note reminders and scribbles, and the report needs to be written before you go home — with the facts intact and nothing invented. | field records daily-report |
| RFI drafting that gets answered | Something on site does not match the drawings and you need an RFI that is clear enough to be answered first time, with the schedule impact stated. | rfi documents coordination |
Task-specific safety documents, written with a competent person.
| Prompt | Use it when | Tags |
|---|---|---|
| Job safety analysis from a task description | You are preparing a task-specific safety analysis and want a structured first draft to edit with the crew, rather than starting from a template that says nothing about this job. | safety jsha field |
| Toolbox talk from a real near miss | Something nearly went wrong on this site and you want a five-minute talk that lands with the crew, rather than a generic briefing they have heard ten times. | safety training briefing |
Notices, risk reviews and payment narratives, drafted for review.
| Prompt | Use it when | Tags |
|---|---|---|
| Contract risk review preparation | A contract or subcontract is on your desk and you want the risk topics organised before it goes to your lawyer, so the expensive hours are spent on judgement rather than on reading out loud. | contracts risk commercial |
| Notice letter draft | An event has occurred and the contract requires written notice. | notices claims correspondence |
Case studies, proposals and the writing that wins the next job.
| Prompt | Use it when | Tags |
|---|---|---|
| Project case study from the facts | You have finished a job worth writing about and need a case study for the website, a proposal or a prequalification pack — without inventing anything and without breaching confidentiality. | marketing case-study bd |
| Tailoring a proposal to the client's actual concerns | You have a standard proposal structure and a specific client, and you want the document to answer what they are worried about instead of repeating your company history. | proposals bidding bd |
For the person who would rather script it than retype it.
| Prompt | Use it when | Tags |
|---|---|---|
| IFC data extraction script | You need a repeatable way to pull data out of IFC models — quantities, spaces, properties — instead of clicking through them, and you would rather review code than write the boilerplate. | ifc bim python automation |
| Job cost analysis in Python or pandas | Your accounting system will give you a job cost export but not the analysis you want, and the work is repetitive enough to script rather than rebuild in a spreadsheet every month. | job-cost python analysis automation |
| Section | What it is for |
|---|---|
| Use it when | The specific moment this prompt earns its place, in one or two sentences |
| The prompt | The text to paste, with {{variables}} for your facts and the output format built in |
| What good output looks like | How to tell a useful answer from a fluent one |
| Follow-ups | Two or three refinements that turn a first draft into something usable |
| Guardrails | What never goes into the tool, and what a person must verify afterwards |
Two habits make the difference between these working and not. Give the model the numbers — a prompt asking for a "sanity check on our estimate" gets generic advice; a prompt with your unit rates, quantities and basis gets an argument you can test. And ask for the counter-case, because a model asked to justify a number will justify it, whether or not it deserves it.
All prompts are written to work on any of the three. Where behaviour differs:
| Situation | ChatGPT | Claude | Gemini |
|---|---|---|---|
| Long documents (specs, contracts) | Good, watch the context limit | Strong: handles long, dense text well | Good, generous context |
| Structured output (tables, schedules) | Ask explicitly for a table or JSON | Follows a stated schema closely | Ask explicitly, then confirm the format |
| Arithmetic and unit conversions | Verify every number: it will do the maths confidently | Better at showing its working | Verify every number |
| Images and renders | Strong for image generation and critique | Text-side critique is strong; image generation is separate | Good image generation, watch style drift |
| Tone for client-facing writing | Can drift to marketing | Usually plainer and easier to edit | Can drift to marketing |
The rule that covers all three: the model drafts, a person decides. Nothing here is a substitute for an estimator's judgement, an engineer's calculation, a competent person's assessment, or your lawyer's reading of a contract.
- Never paste contracts under negotiation, pricing or bid strategy, personal or medical data, anything under an NDA, or anything a client has asked you to protect.
- Always verify quantities, dates, clause references, code sections and any figure that will leave the company.
- Use a business tier with terms that exclude your data from model training, and a written company rule on what may be pasted.
- A person owns every output. If nobody's name is on it, it is not ready to send.
Add a file under prompts/<section>/ with the same five sections, then run:
python3 scripts/build_index.py # regenerate the index and badge
python3 scripts/validate.py # check structure consistencyA useful contribution says where the prompt came from — a real job, a real failure mode — and what it is bad at. Prompts that only sound impressive do not belong here.
Prompts and documentation: CC BY 4.0. Use them, adapt them, credit the source.
This is general material, not advice for a specific project. Nothing here replaces a competent person on site, a qualified engineer's calculation, your contracts, or your lawyer, broker and insurer.
Open construction resources from the same team, all maintained alongside this one:
| Repository | What it is |
|---|---|
| AI in construction | A practical guide to AI in construction: what works today, what the data has to look like, the risks, and a 90-day plan. |
| Construction data migration | A guide and toolkit for moving a contractor between systems, and proving nothing was lost. |
| Construction project records | Open schemas, templates and a checker for RFIs, submittals, change events, daily reports and punch lists. |
| Construction reference data | Cost codes, units, waste factors, pay units, trade sequence, glossary and metric formulas in CSV. |
| Construction reference MCP server | An offline MCP server that gives AI assistants construction reference data and calculators. |
| Construction agent skills | 28 installable agent skills for Claude Code and any agent that reads SKILL.md. |
| Open source construction tools | Open source software for BIM, CAD, scheduling and site work, verified against the GitHub API. |
Maintained by Constructelligence — building the AI infrastructure for construction.