For civil and structural engineers who write up every site visit
You left site at 3. The report is still open at 9.
Writing takes about 40% of an engineer’s working time (VTechWorks): site reports, RFI answers, minutes, fee proposals. None of it gets stamped, much of it can’t be billed, and it gets done in the evening, after the engineering you were hired for.
The Engineer’s Paperwork Pack: one Claude Skill with 32 ready-made playbooks.
- Site observation reports
- RFI responses
- Fee proposals
- One-time payment
- 30-day money-back guarantee
- Works on free Claude
Field Observation Report FOR-007 · excerpt
Purpose and scope. Observation visit on 14 October to review Level 2 composite deck and the column B/2 base ahead of the proposed Level 2 pour. Visual observation of accessible areas only; this report is not an inspection or acceptance of the work.
Observation 1 (Photos 03–05). Shear studs between gridlines C/3 and C/5 appear to be spaced at approximately 16 in. on centre. Drawing S-301, Detail 4 shows 12 in. on centre. Action: Contractor to measure and confirm stud spacing and submit a correction proposal before the deck is poured. Status: open.
- Status
- Draft
- Rev
- A
- Drawn
- Claude
- Checked
- You
Checked before you see it
- Checked: Every observation traced to your notes and photo numbers (01–09)
- Checked: Drawing and spec references taken only from your uploads (S-301 Det. 4; 05 12 00 p. 14)
- Verify: VERIFY: anchor-rod washer requirement, clause not confirmed. Check AISC 360-16 in your licensed copy
Sound like your week?
- You’re back in the car with nine photos and a page of scribbles, and the contractor already wants an answer on Friday’s pour.
- The RFI log keeps growing, and every answer means digging through the spec book for a section you know is in there somewhere.
- A fee proposal is due tomorrow, so tonight you’re copying the last one and rewriting scope, assumptions and exclusions by hand.
- Monday’s site meeting ended with six actions and no minutes. By Thursday nobody remembers who owns what, and the email chain says it was you.
Where the hours go
About 16 hours of your week go on writing
The same week with Claude and the playbooks: about 7 hours back
- Today
- With The Engineer’s Paperwork Pack
Client letters, emails and RFI responses
Today: 6 h/wkWith The Engineer’s Paperwork Pack: 4 h/wkSite observation and technical reports
Today: 4 h/wkWith The Engineer’s Paperwork Pack: 2 h/wkFee proposals and RFP responses
Today: 4 h/wkWith The Engineer’s Paperwork Pack: 2 h/wkMeeting minutes and action logs
Today: 2 h/wkWith The Engineer’s Paperwork Pack: 1 h/wk
Today: ≈16 h a week, ≈768 h a year on this work.
Back in your week, our estimate
≈7 h
Example starting value: $120. Move it to your own number.
- Pro pays for itself after
- ≈30 min
- Time back, each month
- ≈$3,360
How we estimated these numbers
One survey of professional engineers found that about 40% of their time goes on writing, with correspondence the biggest share, followed by technical reports and proposals in roughly equal parts (VTechWorks, an older academic study). On a 40-hour week that’s about 16 hours. No one has published how those hours split by document, so the split below is our estimate.
- Client letters, emails and RFI responses, 6 h/week: Estimate: correspondence is the largest share of the ~16 writing hours (VTechWorks), and client communication and approvals is the #1 non-billable time sink (Total Synergy)
- Site observation and technical reports, 4 h/week: Estimate: reports and proposals take roughly equal shares after correspondence (VTechWorks); the hours split is ours
- Fee proposals and RFP responses, 4 h/week: Estimate: reports and proposals take roughly equal shares after correspondence (VTechWorks); the hours split is ours
- Meeting minutes and action logs, 2 h/week: Estimate: the rest of the ~16 writing hours; no published figure
These are our estimates, not measured results. Hours today split a 40-hour week using the published ~40% writing share (VTechWorks). With the playbooks, we assume half of today’s time is still needed on reports, proposals and minutes, and two thirds on correspondence, where more of the work is judgement and short replies. What’s left is yours: checking every [VERIFY], the engineering decisions and your sign-off. Your own numbers will differ.
Week: A standard 40-hour full-time week, our assumption: we found no published working-week figure for civil engineers, and many work longer. Yearly figures use 48 working weeks, monthly ones 4.
One Skill · 32 playbooks inside
See what actually comes back
One Skill, installed once. You type the job in plain words, the router opens the right playbook, and this is the document you get back, ready for your review.
The visit written up the same day, RFIs answered from your spec, minutes before anyone forgets.
You ask
“Write up today’s visit to Birch Lane townhomes. Notes and photos 01–12 attached: footing rebar at Bldg 2, sill anchors at Bldg 1.”
You get
A field observation report the same day, every observation tied to your notes and photos.
Playbook: site-visit-report.md · in every plan
civil-engineers/
├─ SKILL.md
the router: reads your request, opens the right playbook
├─ your profile, saved at setup
└─ playbooks/
├─ firm-jurisdiction-profile.md
├─ site-visit-report.md
├─ condition-assessment-report.md
└─ … 29 more
See all 32 playbooks
In every planBasic · $29
- The right code family, edition and units in every document
firm-jurisdiction-profile.md - A site observation report the same day as the visit
site-visit-report.md - Condition assessments for bridges, buildings and culverts in a consistent format
condition-assessment-report.md - Minutes and an action log before everyone leaves the car park
meeting-minutes-action-log.md - RFIs and RFI responses written from the designer’s side
rfi-drafter-responder.md - Technical problems explained so the client actually approves
client-explainer-letter.md - Fee proposals with tight scope, assumptions and exclusions
fee-proposal-rfp-builder.md - A calc report narrative and basis of design around calcs you already ran
calc-report-narrative.md - Transmittals and cover letters in seconds
transmittal-cover-letter.md - Client data checked before it reaches the chat
confidentiality-precheck.md - A verification block and AI-use record on every output
verification-block-ai-log.md
Added in ProPro · $59
- Compliance reviews that never pass off a recommendation as code
code-vs-recommendation-review.md - A submittal compliance matrix against your spec section
submittal-review-matrix.md - Spec sections adapted from your master, inconsistencies flagged
spec-section-editor.md - Spec–drawing conflicts caught before the contractor finds them
spec-drawing-consistency-check.md - That 200-page geotech report, every finding with its page number
long-document-digest.md - Straight answers from a tender pack, with the page to prove it
tender-pack-qa.md - Every RFP requirement matched to your response
rfp-compliance-matrix.md - Project sheets and CVs tailored to each bid
project-sheets-cv-library.md - Client approvals chased without the awkward emails
approval-chaser.md - Out-of-scope work turned into a fee letter, not a favour
additional-services-letter.md - Change order and variation justifications that hold up
change-order-variation-justification.md - Method statements and construction sequences, first draft done
method-statement-sequence.md - A design criteria document with codes and editions from your profile
design-basis-criteria.md - Risk reports and compliance statement drafts, without certification language
risk-report-compliance-statement.md - That inherited, uncommented spreadsheet, finally documented
excel-calc-sheet-auditor.md - Your inputs checked for units and signs in executed code
units-sign-checker.md - An independent hand check, written up
hand-check-summary.md - Scripts and glue code for ETABS, SAP2000 and RFEM
analysis-script-helper.md - “What does the code say?” answered with where to look, never made-up text
standards-lookup-guard.md - A peer reviewer briefed on the critical points
peer-review-prep.md - Graduate drafts that arrive with a senior’s questions attached
early-career-review-mode.md
Why not just ChatGPT?
Why not just use ChatGPT?
A blank chat window doesn’t know your formats, your code family and edition, or where the line sits between a code requirement and a recommendation. It drafts fast, then hands you a second job: checking everything it wrote. Engineers who have tried it describe the same failures.
Generic AI
- It turns a recommendation into a “code requirement”. Engineers reviewing AI compliance checks report this as the most common failure, which means every finding has to be re-investigated (Eng-Tips).
- It answers code questions with confidence, but as one Eng-Tips commenter put it, ChatGPT doesn’t know the codes, how they’re interpreted or how they’re enforced.
- Its numbers can be right in magnitude and wrong in sign: tension read as compression, with the error carried through every step after it (CognitiveFuture, on a 2025 study).
- Chat models read complex drawings unreliably. In one vendor’s takeoff test on a 287-page drawing package, Claude priced the job 39% below a chartered quantity surveyor and ChatGPT 43% below (Civils.ai). That’s why the pack treats anything read off a drawing as a first pass for you to confirm.
The Engineer’s Paperwork Pack
Why this works where a blank chat didn’t
- 1
It never invents a clause
A clause number appears only when it’s quoted from a document you uploaded, with its page. If it isn’t in your documents, the draft says “VERIFY: clause not confirmed”. In Pro, compliance reviews keep Code requirement, Recommended practice and Engineer to decide in separate columns.
- 2
Every fact points back to your page
Site reports trace each observation to your notes and photo numbers. Digests, tender Q&A and submittal matrices tie each finding to its document, page and section, so you check a line against its source instead of re-investigating every finding.
- 3
You keep the stamp
Every draft ends with “DRAFT, for review by [name, licence]”, the checks you must make and an AI-use log line. No seal, certification or “complies with code” wording. Your calcs stay in your software: the pack writes the narrative around them and never redoes them.
How it works
Set up once. Then just ask.
No coding, no prompts to copy. You install it the way you would add a document to Claude.
- 1
Download the ZIP
Instant download after checkout: one Claude Skill containing every playbook in your plan, each with its instructions, checklists and output templates.
- 2
Upload it to Claude
Upload it once in Claude’s settings. A router inside reads each request and opens the right playbook, so you never pick a file yourself. It works on Claude Free with code execution turned on, and on any paid plan.
- 3
Set up your firm and jurisdiction profile
A five-minute interview: country and state or province, code family and edition, units, discipline, professional body and signature block. Every playbook reads it from then on.
Compare your options
What the paperwork costs you today
You already have options for the writing around your engineering. Here is what each one costs, and where it falls short.
- Structural software subscription (ClearCalcs, SkyCiv)About $35–119 a month for ClearCalcs; $69–109 a month for SkyCiv’s paid plans (G2)Built for the calculations, which you should keep. It doesn’t write the report, RFI or fee proposal around them, and it bills every month.
- A freelancer or technical writerQuoted per job, every jobEach job needs a brief and a review, you still sign the result, and your client’s project data leaves the firm.
- Your own eveningsHours you can’t billWriting already takes about 40% of an engineer’s time (VTechWorks), about 16 hours of a 40-hour week. Done after hours, it never reaches an invoice.
- Generic ChatGPTFree, or a seat your firm may already pay forNo firm formats and no clause guard. It can turn a recommendation into a “code requirement”, so you check every line (Eng-Tips).
- The Engineer’s Paperwork Pack$29 or $59, onceKeep your calc software for the numbers. For everything you write around them: $29 or $59, paid once, with 30 days to ask for a refund.
Pick your plan
One payment. Yours to keep.
Both plans cost less than one month of SkyCiv Basic ($69, per G2). Pay once, keep it.
Pro costs, once
$59
Basic: $29 once
It pays for itself after
≈30 min
of time won back, at $120 for an engineer’s hour
Get back just 1 hour a month
$120
a month, against a one-time $59. Our estimate is ≈7 h a week.
Break-even = plan price ÷ the hourly value set in “What those hours are worth” above.
Paperwork Core
This week’s reports, RFIs, minutes and proposals, drafted and ready for your review.
$29one-time
11 ready-made playbooks
Try it on real work. Not for you? Full refund within 30 days.
- One Claude Skill with 11 playbooks, including your firm & jurisdiction profile
- Site observation and condition assessment reports
- Meeting minutes with action log
- RFIs and RFI responses from the designer’s side
- Client letters, fee proposals, transmittals
- Calc report narrative around calcs you already ran
- Confidentiality pre-check, verification block and AI-use log line on every output
Paperwork Pro
Long reviews too: submittals, specs, tenders and geotech reports, every finding linked to its page.
$59one-time
All 32 playbooks
Only $30 more than Basic: +21 playbooks and the Pro bonuses below.
Try it on real work. Not for you? Full refund within 30 days.
- Everything in Paperwork Core, plus 21 playbooks (32 in total)
- Code vs recommendation reviews in separate columns
- Submittal review matrices and spec section editing
- Long-document digests and tender pack Q&A with page references
- RFP compliance matrices, project sheets and CVs, approval chasers
- Change order / variation justifications and method statements
- Design criteria, risk reports and compliance statement drafts
- Excel calc sheet audits, units & sign checks, hand-check summaries
- Analysis software script helper, standards lookup guard, peer review prep, early-career review mode
- Pre-filled profiles for the US, UK, Canada, Australia and New Zealand (code family and body names only, no code text)
- Worked examples: a site report, an RFI and a calc narrative, start to finish
- Library of downloadable sample outputs
30-day money-back guarantee
Try it on your next site visit, RFI and fee proposal. If you decide it’s not for you within 30 days, email us for a full refund. No conditions, no forms.
- One-time payment, no subscription
- Works on free Claude
No Claude account yet?
And if I don’t have Claude?
Claude is Anthropic’s AI assistant. The free plan is enough to run The Engineer’s Paperwork Pack: creating an account takes a minute, and you don’t need a credit card.
- Free plan works, no subscription needed to start
- Account ready immediately
- Paid plans only add more usage, for when you want it
- 1
Create a free account
Go to claude.ai and sign up with your email, Google or Apple account. It’s free and ready right away.
- 2
Turn on one setting
In Settings → Capabilities, switch on “Code execution and file creation”. Skills need it, and it’s free.
- 3
Upload your package
In Customize → Skills, click “+”, choose “Upload a skill” and select the ZIP you received. Then just ask.
Fair questions
What civil and structural engineers ask before buying
Does it do my calcs, and who is liable for what it writes?
It doesn’t do your calcs. Analysis, load paths, member design and code checks stay with you and your software. The pack writes around them (calc narrative and basis of design, plus hand-check summaries and units and sign checks in executed code in Pro), and never produces design numbers for you to stamp. Liability stays where it is today: with the licensed engineer. That’s why every output is marked “DRAFT, for review by [name, licence]” and lists what you must check.
Is my board OK with me using AI?
The guidance we’ve read (ASCE Policy Statement 573, the NC Board guidelines, the EGBC advisory adopted by PEO, Engineering New Zealand) says the same thing: you can use AI as long as you keep responsible charge, check the output and can reproduce the result without it. The pack is built around that: drafts for your review, page references, hand-check summaries and an AI-use record as you work. It isn’t endorsed by any board, and your judgement stays yours.
What about client confidentiality?
Every playbook runs a confidentiality pre-check first. It offers to redact client names, site addresses and commercial details, and reminds you about client consent, your firm’s policy and your Claude plan’s training settings. If your firm bans unapproved AI tools, follow your firm’s policy.
Does it include the codes?
No, and that’s deliberate. Standards are copyrighted. Engineering New Zealand warns that uploading NZS text without permission breaches Standards NZ copyright. The playbooks work from your licensed copies and the documents you upload, and point you to where to check.
Built for your rules
Built for the codes and contracts you actually work under
Your profile sets the jurisdiction once. The playbooks then use the right vocabulary, units and record-keeping, and only cite the code family and edition you declared. Pick a jurisdiction to see what changes.
Your setup
United States · IBC / ASCE 7
ASCE 7, ACI 318 and AISC 360 under your adopted building code, imperial units, state-board records.
- Profile records your code family and edition (e.g. ASCE 7, ACI 318, AISC 360) and your state board. Clauses are cited only from documents you upload.
- Imperial units by default, with explicit conversions when a product sheet or test certificate is metric.
- “Change order” wording, and spec sections adapted from your MasterFormat master.
- AI-use log fields follow the NC Board’s record-keeping guidance: inputs, outputs, validation and reviewer.
Guardrails
Guardrails you can see in every output
Every playbook, in both plans, shares the same verification backbone. You see it in every draft: the VERIFY column, the page reference, the review block. The licensed engineer stays in responsible charge, always.
Sample verification block
From: Field Observation Report FOR-007 · excerpt
- Checked: Every observation traced to your notes and photo numbers (01–09)
- Checked: Drawing and spec references taken only from your uploads (S-301 Det. 4; 05 12 00 p. 14)
- Verify: VERIFY: anchor-rod washer requirement, clause not confirmed. Check AISC 360-16 in your licensed copy
- Verify: 16 in. vs 12 in. is a field estimate from your notes; confirm by measurement
- 01
Clauses only from your documents
A clause number appears only when it’s quoted from a document you uploaded or pasted, with document and page. Otherwise the output says “VERIFY: clause not confirmed”.
- 02
Code kept apart from recommendation
Reviews use separate columns: Code requirement (with source), Recommended practice, and Engineer to decide.
- 03
Numbers run as code, with units
Any number goes through executed code (Python or Excel formulas) with explicit units, a declared sign convention, an input table and a hand-check summary. It’s presented as a check of your inputs, never as design.
- 04
Page references on every finding
Every finding or summary point links back to its document, page and section.
- 05
Jurisdiction and edition declared
Read from your profile. If they’re missing, the playbook stops and asks instead of guessing.
- 06
Confidentiality check before input
Offers to redact client names, site addresses and commercial data, and reminds you about client consent, firm policy and your plan’s training settings.
- 07
“Draft for engineer review”, never a seal
Every output ends with “DRAFT, for review by [name, licence]”, mandatory human checks and space for your signature. No stamp, certification or “complies with code” language.
- 08
An AI-use record as you work
A log line on each deliverable: date, tool and model version, inputs supplied, outputs checked, how and by whom.
- 09
Assumptions table for anything read off a drawing
Quantities and drawing interpretations are labelled “first pass” with every assumption listed.
- 10
No code text in the pack
The pack contains workflows, checklists and pointers to your licensed copies, never copyrighted standards text.
FAQ
Everything else
Didn’t find your answer? Email us:hello@summonskills.com
Who is liable for what it produces?
You are, as with any draft a junior or a tool gives you. Every output is marked “DRAFT, for review by [name, licence]”, lists what you must check and never uses seal or certification language. The technical responsibility stays with the licensed engineer.
Can it calculate anything at all?
It runs checks on numbers you give it: unit conversions, sign conventions, a hand check on one case of an inherited spreadsheet. All of it runs as executed code with explicit units and an input table. It won’t size members or run analysis, and quantities read off drawings are labelled first pass with every assumption listed.
Does it work on the free version of Claude?
Yes. The Skill works on Claude Free as long as code execution is turned on in your settings, and on any paid plan. The three setup steps are on this page, under “And if I don’t have Claude?”.
Can I use it with ChatGPT or Copilot?
Not as it’s built. It’s a Claude Skill: you install it once and Claude opens the right playbook for each request, and ChatGPT and Copilot have no equivalent way to install it like that. Claude’s free plan is enough to run it, an account takes about a minute, and you can keep using ChatGPT or Copilot for everything else.
What’s actually in the ZIP?
One Claude Skill. Inside, a router file (SKILL.md) reads what you ask and opens the right playbook, so you never pick a file yourself. The playbooks hold the instructions Claude follows, checklists, output templates and the firm & jurisdiction profile interview. Pro adds pre-filled profiles for five countries inside the Skill, plus a bonus pack with worked examples and sample outputs. There is no copyrighted code text anywhere in it.
Will it read my drawings?
It can read references from drawings you upload, but we don’t promise reliable drawing interpretation. Anything read off a drawing is treated as an input for you to confirm, and labelled first pass with its assumptions.
Imperial or metric?
Whichever your profile says. Units are carried explicitly through every check, and conversions between imperial and metric are shown, not assumed.
What if it isn’t for me?
You have 30 days. If it doesn’t earn its place in your week, email us and we’ll refund you in full, no conditions.
Hand Claude the paperwork. Keep the stamp.
Our estimate: about 7 hours a week back, 336 hours over a 48-week working year. Site reports, RFIs, minutes and fee proposals stop waiting for your evenings: drafted from your own notes and documents, every source on the page, your name on the review block. Try it on your next site visit. If it isn’t for you within 30 days, email us for a full refund.
One payment, yours to keep · 30-day money-back guarantee