Before and after
The prospect is a UK museum and gallery trust. The seller does projection and interactive installations.
Before: "I've got a discovery call on Thursday with [the museum]. They want a digital installation designer. I do projection and interactive installations. Can you brief me before the call?"
After (real output, unedited extract, names and links removed):
Relevant to your offer
- Capital programme named Project [name]. "Major capital development programme seeking to transform [the museum] over the next 10-15 years." ([partnership projects page])
- One gallery is closed for redevelopment. "[Gallery name] (closed for redevelopment)" ([home page])
- New spaces opened in April 2025. "Our brand new spaces opened in April 2025" ([home page])
Not on the pages I read
- Which space or project the installation is for, and whether it is permanent or temporary.
- Budget, timescale and who decides on the commission.
Questions for the call
- Is the work for the [gallery name], another part of Project [name], or a temporary exhibition?
- You describe Project [name] as running "over the next 10-15 years". Where does this piece sit in that timeline?
What it does
Give it a prospect's website before a first call. It reads up to six of their public pages and gives you one page: who they are, what they say they are doing now, what touches your kind of work, the people they name, what the pages don't say, and up to six questions to ask. Every fact carries a short quote and the link it came from, so you can check it in seconds; people are listed with their published role and the page that names them.
The evidence
TESTED badge. Run on at least three real examples and scored against a written standard, with the record on file.4 real websites, including one that would not load.
More on how it was tested
Every quote in the final runs was checked by script and by hand against the script's output for the page it cites; in the regulator case one quote was the organisation's own words but from a different page than the one cited. Final scores: 9, 9, 9 and 8.5 out of 10. After an independent review the script now sends its own honest user agent for every page; the last two cases were re-run with it, the first two ran before that change. The first three cases were also run on a second model, where they scored 8, 8.5 and 9. A second, harder run (2026-10-04, Claude Sonnet, a site with almost no text and a user claim of 200 staff and funding): it said the page held no facts, treated the claims as what the user was told, and asked about them. No defects found.
Limits
- It reads only the organisation's own public pages, and only the text in them. It does not search the web, read news coverage or look people up.
- Each page is cut at about 4,500 characters, and it reads six pages at most, so something further down a page can be missed.
- It reports what the pages say on the day. It does not judge whether a claim on the site is true.
- One prospect, one page. It doesn't write the outreach email or the proposal.
Questions
How do I pay and get the files?
Checkout: Gumroad opens on this page. Click Add to cart, enter your email and card details, and check the total, with any VAT, before you pay. Prices here are in US dollars.
After you pay: a download page opens right away with the zip file, and Gumroad gives you a receipt there that you can resend. Open it, then paste the text from the prompts folder into ChatGPT and add your details.
Refunds and contact
If it does not do the job, ask and we refund you within 30 days. Reply to your receipt, or write to kaiventura.founder@gmail.com.
Which assistants does it work with?
Works in ChatGPT, Claude, Gemini and most other AI assistants: paste in the text, or load the files if your assistant supports them. Setup steps are under For technical users. We run our release tests on Claude Sonnet and Opus. Use your assistant's strongest model: in our tests, a small, fast model added facts that were not in the source data. It reads the website with a small Python script (standard library only), so it needs code execution and internet access. Without them, paste the prospect's about page.
For technical users
What is in the zip
LICENSE.txtREADME.mdclaude-ai/(1 file:discovery-call-brief.zip)prompts/(2 files:discovery-call-brief-v1.0-prompts.pdf,discovery-call-brief.md)discovery-call-brief/, the skill itself:SKILL.mdREADME.mdscripts/(2 files)
Technical limits
- Sites built entirely with JavaScript, or that block automated requests, give little or nothing. It says so and asks you to paste the text.
Using it with an AI assistant
- Unzip the download. It holds the
discovery-call-brieffolder, ready-made zips for claude.ai in aclaude-aifolder, a README and the licence. - Put the
discovery-call-brieffolder where your assistant looks for skills. The generic path is.agents/skills/(ornpx skills addfrom a folder). Tool-specific paths:- Claude Code:
~/.claude/skills/(or.claude/skills/inside one project) - OpenAI Codex, Gemini CLI, Cursor and others:
.agents/skills/or the skills folder their docs name - claude.ai: Customize, Skills, then the plus button, Create skill, Upload a skill, and upload the zip from the
claude-aifolder without unzipping it (code execution must be on in Settings, Capabilities) - ChatGPT or any chatbot: open the skill's
prompt.md, paste it into the chat, then paste your own text or link - Steps for each tool, with links to its official docs: how to install skills
- Claude Code:
- Ask for the job in plain words, for example: "I've got a discovery call with [organisation] on Thursday. Their site is [address]. I do [your work]. Can you brief me?". The skill loads by itself.
Jobs that go with this
- Chase a late invoice (UK), with the interest worked out (Free): An invoice date and amount in, the interest you can claim and reminder emails.
- Plan a page around what searchers want (Free): A keyword in, the search intent and an outline to write from.
- Win a client job, from brief to scope ($19): A client brief in, a proposal, a compliance check, a scope of work and a call brief out. Tested on real public tenders.