prospect-site-diagnostic
Skill1 starsChanged 18 days agoMITView on GitHub
Use when a prospect sends their website link and asks for a call, or replies to a proposal asking for a meeting. Turns that reply into a pre-meeting diagnostic document built on real measurements, plus a short message that books the call. Triggers: the prospect sent their site, prepare a file before the call, audit a prospect's website, reply to a client who asked for a meeting.
In this repository4 other files it gives its agents.
- arabic-humanizerSkill
- keyword-gap-analyzerSkill
- remove-ai-marksSkill
- seo-cannibalization-detectorSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
When a prospect replies to a proposal, sends their website link and asks for a meeting, the output is two things: a short message that …
- Distinctive: line 27
Before any checks, come back to the user with a summary of the prospect's reply: what exactly they said, what they asked for, and where the …
- Distinctive: line 61
Every service type announced on the site
- Distinctive: line 131
Parallel track: if there is one (ads), with "What I can't promise"
- Distinctive: line 180
If the site has a supply side (service provider sign-up, agent accreditation): which is harder for them, attracting customers or attracting …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Prospect Site Diagnostic Before the Meeting
When a prospect replies to a proposal, sends their website link and asks for a meeting, the output is two things: a short message that locks in the time, and a diagnostic document built on real measurements, not on impressions.
The goal is not a full report. The goal is three or four specific observations the prospect can check for themselves in a minute. That is what sets you apart from everyone else who applied to the same job.
Governing Rules
- Every number in the document is measured from the site or from a tool. No numbers from memory.
- Anything you can't prove goes in the "What we need data to determine" section at the end of the document, instead of being stated as fact.
- The message is short and the document carries the details. The message is not the place for the report.
- Never mention developers or the site's technical stack in the document. If the recommendations need new pages, say "requires technical support on your side" and leave it there.
- No personal names and no agency name in the document.
- No em dashes or en dashes.
Step 0: Summary of Understanding
Before any checks, come back to the user with a summary of the prospect's reply: what exactly they said, what they asked for, and where the conversation stands (proposal sent, waiting for a proposal, meeting requested). Get the user's confirmation before you start.
Step 1: Scan the Site Structure
Pull directly from the site with fetch or from the browser:
- sitemap.xml: total number of URLs, and how they break down (core pages, articles, products, language versions)
- robots.txt: clean, or is something blocked
- Homepage: H1, H2 and H3 headings, menu links, hreflang, schema, word count
- Service or product types announced in the copy: these reveal page gaps
- Search the whole homepage copy for any mention of price or cost
The key question here: does every announced service have its own page? A missing page for an existing service is the strongest gap you will find, and the easiest one for the prospect to verify.
Step 2: Measure a Sample of Pages
Open 12 to 16 pages that cover the full range, and measure for each one:
- Word count
- Number of H2 and H3 headings
- Whether there is video or images
- Title, meta description and URL
This table is the hard evidence nobody can argue with. Sort it by word count, ascending, so the gap shows.
What to look for: a repeated template, pages that are just a wrapper for a video, zero H3 headings, a title or URL that doesn't carry a query.
Step 3: Demand Numbers
Pull a list of 25 to 40 keywords from a keyword tool (WriterZen Keyword Planner or whatever is available) covering:
- The main commercial query and its variants
- Every service type announced on the site
- Price, cost and booking
- The informational cluster the site already writes about
Take the clusters and the volume of each, not individual keywords.
Two warnings:
- Don't add up two variants of the same query ("best pizza cairo" and "best pizza in cairo"). Google serves them with the same ranking, so adding them is double counting, not extra demand.
- Qualify difficulty with its scale. KD on an ads scale says bidding is cheap, not that organic ranking is easy. Don't say "low competition" without looking at page one.
Step 4: Search Results from the Prospect's Country (Mandatory)
This is the most important step. Don't skip it.
The search results you see from your location are completely different from what the prospect sees. Use Nightwatch Search Simulator (a browser extension) with Country = the prospect's country and Language = their language.
From page one, extract:
| Item | Why it matters |
| Every result and its type | Government, nonprofit, commercial platform, app store, social content |
| Any price shown inside a search result | If competitors show their prices and the prospect doesn't, that is the strongest observation in the document |
| Any feature a competitor announces in its result snippet | If it announces the prospect's own feature, that is a hit |
| "People also ask" (PAA) questions | Evidence from Google of the gaps |
| Related searches at the bottom of the page | Same thing |
| Whether there is an AI Overview and what its sources are | If a competitor is the source, that is a strong point |
| Whether the prospect's site shows up at all |
If the SERP data from the tool differs from live Google, go with Google and write that caveat in the document header.
Step 5: Check Ads
Open the Google Ads Transparency Centre and search By domain for the prospect's domain and for every commercial competitor on page one, with the region set to the prospect's country:
https://adstransparency.google.com/search?region=<CC>&query=<domain>
This answers: is the prospect advertising? Are competitors advertising? And if the space is empty, that is an opportunity to put forward as a parallel track.
And state the other possibility honestly: an empty ad space may mean nobody has found it, or it may mean people tried and the conversions didn't work. Suggest a limited test campaign before any scaling.
Step 6: Quick Indexing Check
One site:domain search and paging through the result pages, not a separate search for every URL (that gets you blocked after about 30 searches).
Extract: the URLs that show up, whether there are subdomains or language versions, and the "we have omitted some entries very similar" message if it appears.
That "very similar entries" message is gold-standard evidence of a repeated template, and it comes from Google itself.
Mandatory caveat: the number of site: results is not the real indexing figure. Write the pattern, not the number, and put the exact number in the missing data section.
Step 7: Prioritise
Rank by Impact × Confidence ÷ Effort, and order items by how close they are to a sale, not by search volume.
A gap with proven demand, an existing feature and a missing page is always number one.
Building the Document
This order is tested:
- Header: audit date, what was checked, source of the demand numbers, source of the search results data with caveats
- Summary: the findings in four paragraphs, not the methodology
- Demand table: the clusters and their volumes
- Observations (3 to 5), each with a heading that states its finding
- What I found working well: builds credibility before the criticism
- The opportunity: the page one table with prices, and seasonality
- Where the work starts: numbered items, and one line about the technical support needed
- Parallel track if there is one (ads), with "What I can't promise"
- What we need data to determine: a table of each item against what it would settle
Writing Rules
- Observation headings state the finding, not "Observation one". Compare: "Observation three: page structure" versus "One template repeated across forty-one pages"
- When you criticise existing work, state its value first. "This does not mean the content was misplaced" turns the message from "your work was wasted" into "it is missing a layer"
- Numbers come with real examples: a specific URL or query, not abstract statistics
Before Delivery
Run a humanizer skill on the document and on the message (arabic-humanizer in this repo when they are in Arabic).
Patterns that show up in this kind of document in particular:
- Negative contrast (X, not Y): it repeats easily as a rhythm. Count it, and if it appears more than twice, clean it up
- Headings that all follow the same template
- Superlatives: "the clearest opportunity", "the most important point", "the biggest opportunity"
- Bold as decoration on every label
Target: both read as clean, with no tells left from the patterns above.
Delivery
The document: a Google Doc built from HTML (add dir="rtl" when the document is in Arabic or another right-to-left language), with headings and tables that have a coloured header row. The available Drive tool only creates new files. It can't edit the content of an existing file, so any revised version = a new doc.
Remind the user to open up sharing (Anyone with the link, Viewer) before sending the link, or to download it as a PDF and upload that directly. With a new client, the PDF is safer.
The message:
- Under 180 words
- Opens with the strongest observation, not the methodology
- Each observation on its own line, with varied lengths
- The free offer with a specific deadline
- Suggested times in the prospect's time zone, outside their country's weekend
- Closes with a real diagnostic question that opens a discussion in the meeting
Questions for the Meeting, Not the Document
Some things can't go on paper, and they are what decide the size of the offer:
- The value of a single booking and the conversion rate (they turn search numbers into revenue numbers)
- Have they tried ads before
- If the site has a supply side (service provider sign-up, agent accreditation): which is harder for them, attracting customers or attracting service providers? The answer to that can change the whole shape of the offer
- Access to Search Console
Made by Hend Ibrahim · hendebrahim.com · More skills: github.com/hendibrahim-ai/claude-skills
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

