Poster & Prompt Library

AR Site Inspector prompt studio

Ready-to-use, copy-paste prompts — vivid safety-orange marketing poster prompts for image generators, plus practical site-defect examples an inspector could paste into the app today to classify a deviation, draft a non-conformance report or set up a project. Tap Copy on any card.

Poster prompts are written for image models (Replicate, Pollinations, Midjourney, DALL·E, Imagen). Append your aspect ratio if the tool needs it (e.g. --ar 4:5). Describe people generically (an inspector, a site engineer) — never use fake human names, fake firm names, fake counts or superlatives.
Honesty note for all copy: AR Site Inspector is AI-assisted decision support, not a certified engineering or safety inspection — a qualified engineer must verify every finding on site, and the AI can miss or misclassify a defect. The severity view (critical/major/minor) works on every device; the AR plan overlay needs a camera and the printed anchor on site, and is a visual alignment aid, not survey-grade georeferencing. Building plans are used as exported images (not parsed as DWG in the browser). Keep all claims factual and never imply the app certifies a structure as safe.
Marketing posters

Vivid, factual poster prompts

Seven detailed visuals for a construction site-inspection product. Each prompt fully describes subject, style, palette, lighting, aspect ratio and where text goes. Confident and factual — an inspector in a hi-vis vest and hard hat holding a tablet that overlays a building plan on the real site via AR with red, amber and green defect markers, a safety-orange aesthetic, no fake names or counts, no overstated claims, and never implying the app certifies a structure as safe.

Hero — AR plan overlay with defect markers4:5

Flagship visual for landing pages & print.

Confident, premium poster for "AR Site Inspector". A construction site inspector wearing a bright hi-vis safety vest and a hard hat holds a rugged tablet up in front of a real, active construction site (a partly built multi-storey concrete structure with scaffolding). On the tablet screen, and appearing to float in the air aligned to the structure, is a semi-transparent building plan overlaid on the real site as an augmented-reality layer, with glowing defect markers pinned to specific spots — some red, some amber, some green — and thin lines linking each marker to a point on the structure. Palette: safety orange #c2410c and warm amber #f59e0b, with the three semantic severity colours (critical red #dc2626, major amber #f59e0b, minor green #16a34a) glowing on the markers, against a natural daytime construction-site background. Bright outdoor daylight mixed with the orange-amber glow of the AR overlay, contemporary premium editorial photography, soft-focus background, 8k, grounded and believable (not sci-fi). Clean negative space at the top for a headline; room for a small logo bottom-left. Describe the person generically as a site inspector, no fake name, no real firm name.

Overlay text idea: "Overlay the plan. See every deviation." · small: "AI decision-support, not a certified inspection · by Archana Info Solutions".
ReplicatePollinationsMidjourney --ar 4:5

Feature — photograph & assess a defect1:1

The photo-compliance entry point.

Square poster showing a close-up of a site inspector's gloved hand holding a phone, photographing a concrete column with visible honeycombing on an active construction site, while on the phone screen a clean safety-orange-and-amber UI shows the photo being assessed and an "AI compliance check" button ready. A subtle progress indicator and a severity badge begin to form on screen. Active construction site softly blurred behind, hi-vis orange and amber UI accents, a hard hat just in frame. Bright, focused, professional mood, natural daylight, premium editorial photography, shallow depth of field, 8k. Describe the person generically as a site inspector, no fake name. Clear space at the bottom for a caption.

Overlay text idea: "Photograph the defect. Let AI read the deviation."
ReplicatePollinations--ar 1:1

Feature — the critical/major/minor snag map4:5

The severity-graded snag list.

Vertical 4:5 poster showing a clean tablet screen displaying a "Snag List": a vertical list of snag cards, each with a small coloured severity badge — critical red, major amber or minor green — a snag title, a category chip, an IS-code reference and a location line, with a small severity donut at the top, against a warm off-white background with a subtle safety-orange pattern and amber accents. A site inspector's gloved hand rests at the edge of the tablet, blurred. Modern, professional, trustworthy UI-focused composition, premium 8k render, soft shadows, no clutter. Clean banner space at the top for a headline.

Overlay text idea: "Every snag, flagged critical, major or minor." · keep the "not a certified inspection" note in copy.
ReplicateDALL·E--ar 4:5

Feature — the printed QR site anchor1:1

Print an anchor, align the plan on site.

Square poster showing a printed QR-code anchor label fixed to a concrete wall on a construction site, with a site inspector holding a tablet scanning it; on the tablet screen a semi-transparent building plan begins to snap into alignment over the real wall as an AR overlay. Hard hats and scaffolding softly blurred behind. Safety-orange and amber accents, bright outdoor daylight, premium 8k render, soft glow, shallow depth of field. Clear space at the bottom for a caption. No fake names.

Overlay text idea: "Print the anchor. Align the plan on site." · note: a visual alignment aid, not survey-grade.
ReplicatePollinations--ar 1:1

Feature — the PDF snag report4:5

The geotagged PDF snag report.

Vertical 4:5 poster showing a crisp printed PDF snag report lying on a site-office desk beside a tablet, the report showing a firm header, project name and type, severity counts (critical/major/minor), an executive summary and a snag-by-snag table with small embedded geotagged site photos. A small amber "Download PDF" button glows on the tablet UI above it. Safety-orange and amber accents, warm off-white background, natural daylight from a window, premium 8k editorial photography, shallow depth of field. Clean space at the top for a headline.

Overlay text idea: "A clear snag report, geotagged photos and all, in one tap." · small: "AI decision-support, not a certified inspection".
ReplicatePollinations--ar 4:5

Feature — draft an NCR from a snag1:1

The one-tap non-conformance report.

Square poster showing a tablet screen with a clean "Non-conformance report" UI — a snag summary on the left, and on the right an AI-drafted NCR showing the observation, the deviation, an IS-code reference and a corrective action, with a "Generate NCR" button. Safety-orange and amber accents, a softly blurred active construction site behind, a hard hat in frame. Bright, orderly, premium 8k render, soft glow, shallow depth of field. Clear space at the bottom for a caption.

Overlay text idea: "Turn a snag into a non-conformance report." · an engineer reviews every draft.
ReplicatePollinations--ar 1:1

Brand — plan meets site16:9

An abstract banner for headers & ads.

Wide 16:9 abstract brand banner for "AR Site Inspector". A hard hat and a rolled building plan rendered in a modern, minimal, premium style, with the flat plan lines lifting off the paper and re-forming into a neat 3D grid of small red, amber and green markers that float upward over a faint outline of a construction structure, like an AR overlay. Warm off-white background with safety-orange #c2410c and amber #f59e0b linework and a soft glow, the three severity colours (critical red #dc2626, major amber #f59e0b, minor green #16a34a) as accent points. Elegant, industrial, editorial, 8k, plenty of clean space on the left for a headline and a small logo. No text baked in, no people, no fake names.

Overlay text idea: "Where the plan meets the site." · keep the honesty note in surrounding copy.
ReplicatePollinations--ar 16:9
Practical inputs

Sample defects & site text an inspector could paste today

Real starting points to paste straight into the app — a described construction defect to classify with a deviation, severity and likely IS-code, a snag to draft a non-conformance report for, and project types to inspect. These are inputs YOU provide; the AI reasons only over the details given, and its output is a review aid a qualified engineer must verify on site — never a certified inspection.

Defect to classify — concrete honeycombingClassify

A structural defect for the deviation classifier.

On a residential RCC column at the 3rd-floor slab junction there is visible honeycombing exposing the aggregate over roughly a 200 mm x 150 mm area on the north face, with a shallow depth exposing the outer reinforcement in one spot. The surrounding concrete appears sound. Classify the deviation type and severity, and reference the most likely relevant Indian Standard, then suggest a corrective approach.

Use for: the "AI compliance check" flow on ASI_SnagCapture.aspx (or as a text description of a defect) — it returns the deviation type, severity, a likely IS-code and a recommendation. A qualified engineer must verify it on site.

Defect to classify — unsafe temporary wiringClassify

A safety deviation for the classifier.

On a commercial site, temporary electrical wiring is run loosely across a walkway at floor level with exposed joints taped rather than in proper connectors, no earthing visible, and a distribution board left open next to a wet area. Workers cross this walkway regularly. Classify the deviation type and severity, reference the most likely relevant safety standard, and suggest an immediate corrective action.

Use for: the AI compliance check — this should surface a safety deviation at a high severity with an immediate corrective action. Confirm and act on site.

Defect to classify — dimensional deviationClassify

A dimensional/workmanship deviation.

A brick partition wall on the 2nd floor is out of plumb by about 18 mm over its 3 m height and the door opening is 40 mm narrower than the approved plan dimension. The plaster is not yet applied. Classify the deviation type and severity, reference the likely relevant standard for the tolerance, and suggest a corrective approach before finishing.

Use for: the AI compliance check with a dimensional/workmanship defect — it returns the deviation, severity and a likely IS-code tolerance reference. An engineer confirms against the approved drawings.

Snag to draft an NCR for — exposed reinforcementDraft NCR

A saved snag for the NCR drafter.

Snag title: Exposed reinforcement at beam-column junction, Tower B, 4th floor. Description: Honeycombing at the beam-column junction has left a length of the main reinforcement exposed and un-encased in concrete over roughly 250 mm. Cover is inadequate at this location. Deviation type: workmanship. IS-code reference: IS 456:2000. Draft a formal non-conformance report for this snag, with the observation, the non-conformance, the reference standard and a clear corrective action.

Use for: the "Generate NCR" action on ASI_SnagCapture.aspx / ASI_SnagList.aspx — it drafts the NCR text, a corrective action and a reference standard. A qualified engineer reviews the wording before it is issued.

Project types to inspectProject type

Types the app supports for inspection.

Residential Commercial Infrastructure

Use for: pick the closest type on ASI_ProjectSetup.aspx when you set up a project — the type gives the AI useful context when it assesses site photos for that kind of build.

Snag categories to recordCategory

The snag categories the app uses.

Structural Electrical Plumbing Finishing Safety Other

Use for: the category selector on the snag form — the AI pre-fills the most likely category from the photo, and you can adjust it before saving.
Ops & AI system

The working prompts inside the product

The exact system-prompt shapes the app uses (routed through ASI_ClaudeHelper), plus day-to-day operational prompts. Copy into your AI provider or config. Every prompt keeps the strict rule: reason only over the photo or details provided, return strict JSON, and state that the output is decision support a qualified engineer must verify — never a certified inspection.

Photo-compliance vision system prompt

AssessCompliancePhoto — read a site photo.

You are assisting a qualified civil / construction site inspector with a first-pass, photo-based compliance check. This is decision support, NOT a certified engineering or safety inspection. Assess ONLY what is visible in the site photo provided; do not invent conditions you cannot see. Project type: "{{project_type}}". Location on site: "{{location_context}}". Inspector note: "{{inspector_note}}". Return STRICT JSON only, no prose: { "observation": "what you can see in the photo, in plain language", "complianceStatus": "compliant | non-compliant | observation", "deviationType": "material | dimensional | workmanship | safety | none", "severity": "critical | major | minor", "isCodeRef": "the most likely relevant Indian Standard code, or empty", "recommendation": "a clear corrective recommendation", "category": "Structural | Electrical | Plumbing | Finishing | Safety | Other" } If you are unsure, say so in the observation and use "observation" / "minor". Always remind (implicitly, via a conservative reading) that a qualified engineer must verify on site. No emojis, no hype.

Note: Uses the Groq vision model meta-llama/llama-4-scout-17b-16e-instruct. The page sends a proper base64 data image; the returned object pre-fills the snag form and is stored as ComplianceJson.
Groq vision (llama-4-scout)System prompt

Deviation-classify system prompt

ClassifyDeviation — a described defect.

You are assisting a qualified site inspector. Classify a described construction defect. This is decision support, NOT a certified inspection. Reason only over the description provided. Project type: "{{project_type}}". Defect description: {{description}} Return STRICT JSON only, no prose: { "deviationType": "material | dimensional | workmanship | safety | none", "severity": "critical | major | minor", "category": "Structural | Electrical | Plumbing | Finishing | Safety | Other", "isCodeRef": "the most likely relevant Indian Standard code, or empty" } Base it only on the description. A qualified engineer must confirm the finding on site. No emojis, no hype.

Note: Powers a text-only classification when there is no photo; the result feeds the same snag fields.
Groq / Claude APISystem prompt

NCR-drafting system prompt

GenerateNCR — a non-conformance report.

Draft a formal non-conformance report (NCR) for a construction snag. This is a drafting aid, NOT a certified inspection. Snag title: {{snag_title}}. Description: {{snag_description}}. Deviation type: {{deviation_type}}. IS-code reference: {{is_code_ref}}. Return STRICT JSON only, no prose: { "ncrText": "the full NCR text: observation, the non-conformance, and reference to the standard", "correctiveAction": "the recommended corrective action", "referenceStandard": "the relevant standard code, or empty" } Keep it factual and drawn only from the snag details given. A qualified engineer must review the NCR before it is issued. No emojis, no hype.

Note: The NCR text is saved on the snag via ASI_usp_SetSnagNcr; the reference standard is stored back on the snag's IS-code field.
Groq / Claude APISystem prompt

Snag-report summary system prompt

SummariseSnagReport — executive summary.

You are an experienced construction site inspector writing the summary of a site inspection report. You are given the project name and a JSON array of the snags found. This is decision support, NOT a certified inspection. Project name: {{project_name}}. Snags (JSON array): {{snags_json}} Return STRICT JSON only, no prose: { "executiveSummary": "a 2-3 sentence summary of the inspection findings", "keyRisks": [ "a key risk", "another" ], "overallStatus": "satisfactory | needs-attention | critical", "recommendation": "an overall recommendation" } Base the summary only on the snags provided; do not invent snags or figures. A qualified engineer must verify. No emojis, no hype.

Note: The keyRisks value is a top-level JSON array inside the object — parse defensively. The whole object is saved via ASI_usp_SaveSnagReport and reused by the PDF report.
Groq / Claude APISystem prompt

Onboarding script

First-run welcome for a new inspector.

Act as a friendly onboarding guide. Write a short 5-step first-run script welcoming a new inspector to AR Site Inspector: (1) a warm welcome noting the app helps them inspect sites faster with an AR plan overlay and AI photo compliance checks, (2) set up a project, upload its building plan (an image or PDF exported from DWG) and print the QR anchor to place on site, (3) open the AR inspection to overlay the plan on the real site and align it to the printed anchor, (4) capture a snag by photographing a defect so the AI reads the deviation, severity and likely IS-code, then save the snag and optionally draft an NCR, (5) review the snag list and snag report, generate an AI summary and download a PDF with the geotagged photos. Keep each step to 1-2 sentences, confident and jargon-free, and state clearly that the app is AI decision support, not a certified inspection, and that a qualified engineer must verify every finding. No emojis, no hype.

Use for: the in-app first-run walkthrough and welcome email.
ClaudeChatGPT

Demo script

Walk a prospect through the product.

Act as a product specialist. Write a 5-step demo script to show AR Site Inspector to a construction firm or a site engineer: (1) an opening line that the app turns a site walk into a plan-vs-site comparison and a graded snag list, (2) set up a project, upload a building plan and print the QR anchor, (3) open the AR inspection to overlay the plan on the real site and align it to the anchor on supported devices while an orbit view works everywhere, (4) photograph a defect and run the AI compliance check to reveal the deviation, severity, likely IS-code and a recommendation, then save it as a snag and draft an NCR, (5) show the snag list, the severity donut and the AI executive summary, then download the geotagged PDF snag report. Keep each step to 1-2 sentences, practical, confident, no jargon, no hype, and stress throughout that the app is AI decision support, not a certified inspection, and that a qualified engineer verifies every finding.

Use for: presenting to construction firms, project managers and site engineers.
ClaudeChatGPT

Renewal reminder

Gentle per-project renewal nudge.

Write two short renewal reminder messages for AR Site Inspector (one for WhatsApp under 60 words, one for email under 90 words) for the firm "{{firm_name}}" whose plan is ending on {{expiry_date}}. Remind them that keeping the plan active keeps the AR plan overlay, AI photo compliance checks, AI-drafted NCRs and PDF snag reports available for their inspectors; if available, note how many snags were recorded, using only the figures provided. End with a clear, friendly link to renew (the amount, written as "Rs. {{amount}}" per project). Factual and confident, no emojis, no pressure, no fear-mongering, no invented numbers. Indian English.

Use for: the automated renewal reminder before SubscriptionExpiry. Payment via the Razorpay flow on the Subscription page. AR Site Inspector is priced per project.
ClaudeWhatsAppEmail

Social captions pack

LinkedIn + Instagram, construction tone.

Write 8 short social captions for "AR Site Inspector", a tool where an inspector sets up a project, uploads a building plan (used as an exported image), overlays it on the real site in AR aligned to a printed QR anchor (on supported devices) or an orbit view (everywhere), photographs site defects to get an AI photo-compliance assessment flagging the deviation, severity and likely IS-code with a recommendation, builds a critical/major/minor snag list, drafts non-conformance reports, and downloads a geotagged PDF snag report. Mix LinkedIn (practical, construction-operations focused) and Instagram (crisp, professional) styles. Each under 220 characters, one clear value point, 3-4 relevant hashtags, no emojis, no superlatives, no fake counts, never claim the app certifies a structure as safe, and never imply the AR overlay is survey-grade or works on every device. State the AI is decision support that a qualified engineer must verify.

Use for: a week of scheduled posts. Keep every line factual — AI decision support, not a certified inspection; the severity view works everywhere, AR on supported devices.
ClaudeChatGPT