Ask the Dataset What You Didn't Know to Ask
Find the questions your data answers that you haven't thought to pose yet
You extract what you came looking for, then close the file. The patterns you didn't anticipate stay buried because you never asked about them. Six months later someone spots what was always there.
That is the gap this drop closes. It is a data & analysis skill built for Claude, and it takes about ten minutes to set up the first time. After that it runs in under a minute.
Who should use it
Analysts and managers who already use Claude for data summary and want to surface insights beyond their initial brief
How it works
The skill file does five things, in this order.
1. Load the baseline. Give Claude your dataset and a one-sentence description of what it tracks, without stating what you need from it yet
2. Request the anomaly scan. Ask Claude to list anything in the data that contradicts common assumptions in your sector or diverges from expected distribution
3. Isolate silent variables. Identify columns or fields you haven't queried and ask what patterns emerge when you segment by those dimensions
4. Run the null hypothesis. Ask Claude what the data would look like if your current strategy or assumption were wrong, then compare that shape to what you actually see
5. Request the question list. Have Claude generate five questions this dataset could answer that a stakeholder in a different department would care about
What comes back
A working list of non-obvious questions and a scan for patterns you didn't commission, ready to feed back into planning or stakeholder updates
The mistake to avoid
Run this after you've done your primary analysis, not instead of it. The goal is to catch what falls outside your frame, so you need a frame first. Treat any surprise finding as a hypothesis, not a conclusion, and verify it with a second cut of the data or a domain expert.
Running it
Save this as SKILL.md inside a folder named after the skill, then drop the folder into your Claude Code skills directory or upload it to a Claude Project. Claude reads the frontmatter to decide when the skill applies, so keep the description line intact. For a one-off run, paste the prompt block directly.
Where this fits
On its own, one skill saves an hour a week. The compounding happens when three or four of them run in sequence on the same input, the same transcript that produces a scope of work also produces the follow-up email and the project brief. That is the point at which it stops being a prompt and starts being an internal tool. If you want that wired into the systems your team already uses, that is the work 67 Digital does.
In the file
I'm attaching a dataset that tracks {brief description of what the data measures and the time period or scope it covers}.
I've already run my primary analysis. Now I want you to surface patterns and questions I haven't looked for.
First, scan for anomalies: identify anything that contradicts typical assumptions in {your industry or domain} or shows unexpected distribution.
Second, list any columns or variables I haven't mentioned and explain what patterns emerge when you segment the data by those dimensions.
Third, assume my current working assumption is {state your main hypothesis or strategy}. Describe what the data would look like if that assumption were incorrect, then tell me whether the actual data fits that incorrect profile or not.
Fourth, generate five questions this dataset could answer that would matter to someone in {name a different department or stakeholder group}, not just my team.
Finally, if you were presenting this data to someone with no prior briefing, what is the single most surprising or non-obvious pattern you would highlight and why?
Format your response as a numbered list under clear headings. If you need clarification on any variable or column name, ask before proceeding.