A free Claude skill by Alan Weibel

Case study.

Tell Claude about a project you worked on, in your own words. This skill turns it into a case study draft for your portfolio, then asks you for the details you didn't mention, like the numbers.

Download the skill See an example

98%: of the test checks passed with the skill, compared with 66% without it
26 of 26: test requests handled correctly. It starts when you ask for a case study and stays out of the way otherwise
10: sections, based on what hiring managers say they look for
0: made-up facts allowed. When something is missing, it asks you

Overview

Most designers can explain a project well out loud. When they write it down, the numbers, the false starts and the disagreements with other teams tend to get left out, and those are the parts hiring managers look for.

The skill works in three steps.

It writes a draft from what you said. You can talk, dictate or paste notes. It sorts everything into ten sections, writes in the first person, and marks anything it doesn't know with a placeholder instead of guessing.
It asks about what's missing. It rates each section and asks up to five questions at a time, starting with impact: the number before, the number after, the time period, and where you might find them.
It adds the case study to your portfolio. It finds your existing case study page, template or doc, fits the sections into it, and keeps the page marked as coming soon until the gaps are filled.

The structure comes from people who hire designers: a Nielsen Norman Group survey of 204 UX hiring managers, the design hiring guides from Etsy and Google, a former head of product design, a design hiring manager and a former recruiter. Links are in the reference section.

Download

case-study.skill Skill package, 7 files, free to use and share

Install

Claude Code

Unzip it into your skills folder and it will be available in every project.

unzip case-study.skill -d ~/.claude/skills/

You can also unzip it into a single project's .claude/skills/ folder, for example your portfolio repo.

Claude apps

If your plan supports custom skills, upload case-study.skill in the skills settings.

Use

You don't need a special command. Asking for a case study, a portfolio piece, a project write-up, a portfolio entry or a work sample will start it, and so will asking what an existing case study is missing. You can also type /case-study.

Examples

A summary that works

Weak

"I redesigned the onboarding experience for our enterprise product to make it better for users."

Better

"Enterprise admins were giving up halfway through a 14-step setup wizard. I led research and design for a checklist they could pause and come back to, and setup completion went from 41% to 78% in the first quarter."

This example comes from the skill's own guidance, and its numbers are made up. A good summary covers the problem, what you did and the result in two sentences.

Writing about a disagreement

Sales wanted to keep the wizard. Our VP of Sales thought the wizard helped close deals because it looked good in demos. Instead of arguing about taste, I showed her five session recordings of admins quitting at step 9, and suggested keeping a demo mode for sales calls. We gave the checklist to new accounts first so we could compare the two, and once the numbers came in, Sales started using the checklist in their own demos.

Hiring managers say a project where nothing went wrong reads as edited, so the skill gives disagreements and failed ideas their own sections.

How it marks what it doesn't know

METRIC median time from signup to go-live, before and after, over the first 90 days
ASK who pushed back on removing the wizard, and why?
IMAGE the old and new setup screens side by side, with the removed steps circled

You'll see these in the draft. They're hidden on the published page, so a recruiter never sees an unfinished sentence.

A question it asked

"You said setup got way faster and implementation calls dropped a lot. Do you remember roughly how long it took a new customer to go live before and after, and how many implementation calls or hours a typical setup needed before and after, over what period? You mentioned rolling the checklist out to new customers first so you could compare. That comparison is probably the best place to look, along with whatever analytics tool you used, launch or QBR decks, the implementation team's call logs, or asking Dana. If exact numbers are gone, would rough wording like "from weeks to days" be accurate?"

From the example further down the page. It uses details from the notes, asks for every part of the number, and suggests where to find it.

Docs

The ten sections

Section	What goes in it
Summary	A one or two sentence summary of the problem, what I did and the result, plus three key takeaways
Context	Who it's for, what was at stake for the business, why it happened then, constraints and scope
Role	What I owned, what the team did, partners by function
Research	Research methods, the people it was designed for, something a user said, and what the findings changed
Process	The options explored, why one was chosen, and how it was tested
Setbacks	What didn't work, ideas that were dropped, changes of direction
Disagreements	A real disagreement, both sides of it, how it was settled and what was given up
Solution	The final design and the details you cared most about
Impact	Results for the business and for users, with numbers before and after, the time period, and how they were measured
Looking back	What I learned and what I'd do differently

How it behaves

It doesn't make anything up. Every number, quote and result has to come from what you said, and a script checks each number in the draft against your notes.
It checks things you weren't sure about. If you say "I think it was fourteen steps", it uses the number and adds it to a list to confirm with you.
It asks five questions at most per round. You can answer them out loud, in any order.
It keeps your notes private. Your raw notes are saved in .case-study/, away from anything that gets published. That matters if they mention a client under NDA.
It handles NDAs. A named client becomes something like "a large US insurer", and the shot list tells you to recreate screens rather than use screenshots.
It writes like a person. No em dashes, no arrows and no stock phrases, since the case study goes out under your name.
It uses your layout. It fits the ten sections into your own template and saves the result in case-study.config.json, so your next case study goes in without the same questions.

Live output

This is one unedited run from the tests, made with the current version of the skill. The input is a dictated ramble about a made-up project, with no numbers in it. The draft, the scorecard and the questions are exactly what the skill wrote. The highlighted parts are placeholders for facts it didn't have.

1. What was said ramble.txt, 519 words, dictated

okay so this is the fleet onboarding thing at Routewise, I was there from like early 2022 into 2023, I was the lead product designer on the admin side. Routewise does routing and dispatch software for delivery companies, mid size fleets mostly, like regional grocery and parts distributors. So the problem was setting up a new customer was awful. Like the admin had to go through this setup wizard, I think it was fourteen steps, uploading drivers, vehicles, depots, shift rules, all of it, and people would just bail halfway. Our implementation team was basically doing it for them on calls, which is super expensive, and sales was promising go-live in two weeks and it was taking way longer. So the business wanted to cut implementation cost and get customers live faster.

I worked with a PM, Dana, and the onboarding squad, I think six engineers, and we borrowed a researcher part time. I did the interviews myself mostly, we talked to I want to say twelve admins, some new customers mid-setup and some who'd churned. Also I sat in on a bunch of the implementation calls which was eye opening. The big thing we learned was admins don't have all the data at once, the driver list lives with HR, the vehicles with the fleet manager, so they get stuck at step four or five waiting on someone and then the wizard times out or they lose their place. One admin said something like "I feel like I'm doing someone else's homework." That kind of stuck with me.

So the idea was kill the linear wizard and make it a checklist where you can do things in any order, save and come back, and invite a teammate to do their part, like send the vehicle upload to the fleet manager. We also did a CSV importer that tells you what's wrong row by row instead of just failing.

What didn't work, we first tried a smarter wizard with branching and it tested badly, still felt like a tunnel. And the first importer version tried to auto fix data and people didn't trust it, so we switched to showing the problems and letting them fix.

There was a big fight with sales honestly. The sales VP loved the wizard because it demoed really well, it looked guided and fancy. They didn't want us to remove it. I ended up making a demo mode, and we rolled the checklist out to new customers first so we could compare. Engineering also pushed back on the invite-a-teammate thing because of permissions, it touched the roles model, so we scoped it to a limited "task assignee" role first.

Impact, I don't remember the exact numbers, but setup got way faster and implementation calls dropped a lot, and the CS team was really happy. It became the default for all new customers. I think it also helped with a couple big deals.

If I did it again I'd bring sales in earlier instead of fighting them later, and I'd instrument the old wizard before we changed it because we didn't have a good baseline.

2. The draft it wrote case-studies/routewise-onboarding.md

Replacing Routewise's setup wizard with a checklist admins could share with their team

role: Lead Product Designer, admin side
team: 1 PM, onboarding squad of six engineers, 1 part-time researcher
timeline: Early 2022 into 2023
platform: [[ASK: platform, e.g. web admin console?]]
status: shipped

TL;DR

New Routewise customers were abandoning a fourteen-step setup wizard halfway, so our implementation team ended up doing setup for them on calls. As lead product designer on the admin side, I worked on replacing the wizard with a checklist admins could do in any order, save and come back to, and hand parts of to teammates, and it became the default for all new customers. METRIC setup time and implementation calls per customer, before and after

Key takeaways

Result: the checklist became the default setup for every new customer, and setup got faster with fewer implementation calls. METRIC the numbers behind "faster" and "fewer"
Biggest tension: removing a wizard the sales VP liked because it demoed well. I built a demo mode and we rolled the checklist out to new customers first so we could compare.
Skill it shows: basing the redesign on interviews and implementation calls, which showed admins were stuck waiting on HR and fleet managers for data.

Setting up a new customer was slow and expensive

Routewise makes routing and dispatch software for delivery companies, mostly mid-size fleets such as regional grocery and parts distributors. Before a customer could use it, their admin had to go through a setup wizard covering drivers, vehicles, depots, shift rules and everything else.

For admins, the wizard was painful. People would bail halfway through.
For the business, it was expensive. The implementation team was basically doing setup for customers on calls.
For sales, timelines slipped. Sales was promising go-live in two weeks, and it was taking way longer.

The business wanted to cut implementation cost and get customers live faster. ASK why did this get prioritised in 2022, and were there constraints like a deadline or legacy tech?

My role

I was the lead product designer on the admin side. I worked with our PM, the onboarding squad of six engineers, and a researcher we borrowed part time. I ran most of the user interviews myself and sat in on the implementation team's setup calls. I also built the demo mode for sales.

ASK which parts did you personally design (the checklist, the teammate invites, the CSV importer), and what did the PM, researcher and engineers own?

Admins didn't have all the data at once

Who we talked to. We interviewed twelve admins, a mix of new customers in the middle of setup and customers who had churned. I did most of the interviews myself. I also sat in on a lot of implementation calls, which was eye-opening.

What we learned. The admin running setup doesn't own all the information it asks for. The driver list lives with HR, and the vehicle list lives with the fleet manager. So admins got stuck around step four or five waiting on someone else, and then the wizard would time out or they'd lose their place.

One admin put it this way:

"I feel like I'm doing someone else's homework."

That stuck with me.

IMAGE a simple map of who holds which setup data (admin, HR, fleet manager) against the wizard's steps, marking where admins stalled

From a linear wizard to a checklist in any order

So the idea was to get rid of the linear wizard and turn setup into a checklist where admins could:

do the steps in any order
save and come back when they're waiting on someone
invite a teammate to do their part, for example sending the vehicle upload to the fleet manager

Alongside that, we built a CSV importer that tells you what's wrong row by row instead of just failing.

ASK how did you test the checklist before launch, and what did you show people (prototypes, usability sessions)?

IMAGE early checklist sketches or flows next to the old wizard, captioned with what changed

What didn't work

A smarter wizard was still a tunnel. Before the checklist, we tried a smarter wizard with branching. It tested badly, because it still felt like a tunnel.

Auto-fixing data lost people's trust. The first version of the importer tried to fix bad data automatically, and people didn't trust it. We switched to showing the problems and letting admins fix them themselves.

Sales wanted to keep the wizard

Sales. The sales VP loved the wizard because it demoed really well. It looked guided and fancy, and sales didn't want us to remove it. I ended up making a demo mode, and we rolled the checklist out to new customers first so we could compare the two. ASK how did the sales VP respond once the checklist was live, and did sales keep using the demo mode?

Engineering. Engineering pushed back on inviting teammates because of permissions. It touched the roles model. We scoped it to a limited "task assignee" role first.

ASK who proposed the task assignee role, and what did you give up from the original invite idea?

A setup checklist that can be shared

The shipped design replaced the wizard with a checklist that answers each problem from research:

Any order. Admins can work on whatever they have the data for.
Save and come back. Waiting on HR or the fleet manager no longer means starting over.
Assign a task to a teammate. An admin can send a task, like the vehicle upload, to the person who has the data, through a limited "task assignee" role.
A CSV importer that explains itself. It shows what's wrong with each row so admins can fix it, instead of failing or fixing the data for them.

IMAGE the final checklist with a task assigned to a fleet manager, annotated IMAGE the CSV importer showing row-by-row errors, annotated

Impact

It became the default. The checklist became the setup flow for all new customers.

Setup got faster and implementation calls dropped. METRIC time from contract to go-live, or setup completion rate, before and after, with the time period and source METRIC implementation calls or hours per new customer, before and after

The CS team was really happy with it. ASK is there a quote from CS or implementation about the change?

Sales. It may have helped with a couple of big deals. ASK which deals (described generally), and how did the checklist help?

We didn't have a good baseline from the old wizard, so any before numbers will need a note on where they came from.

Looking back

I'd bring sales in earlier instead of fighting them later.

I'd instrument the old wizard before changing it. We didn't have a good baseline, which makes it harder to show how much the checklist helped.

ASK what happened to the checklist after you left, and what from this project have you used since?

To confirm (not for publishing)

The wizard had fourteen steps ("I think it was fourteen").
The squad had six engineers ("I think six").
You interviewed twelve admins ("I want to say twelve").
The exact wording of the admin's quote ("something like").
It helped with a couple of big deals ("I think").

3. What it asked next its reply, scorecard and questions

I turned your Routewise transcript into a draft case study, at case-studies/routewise-onboarding.md. It's about 1,300 words and has all ten sections. The story is there: admins stalling while they waited on HR and the fleet manager, the "doing someone else's homework" quote, the branching wizard that still felt like a tunnel, the importer that auto-fixed data nobody trusted, and the disagreements with sales and engineering. The weak spot is impact. You said setup got "way faster" and calls "dropped a lot", but there aren't any numbers yet.

Some notes on how I wrote it:

I saved your transcript word for word to .case-study/routewise-onboarding.source.md. Your answers will go in that file too. It sits outside the draft folder so it doesn't get published by accident (it has Dana's name in it). This folder isn't a git repo, so I didn't touch a .gitignore.
Every fact comes from what you said. Wherever something is missing there's a METRIC ..., ASK ... or IMAGE ... placeholder: 4 metric, 9 ask and 5 image right now. The draft checker found no numbers that don't trace back to you.
I used "I" for the things you said you did yourself: lead designer on the admin side, most of the interviews, sitting in on implementation calls, and the demo mode. You said "we" for the checklist, the importer and the branching wizard, so I kept "we" there and added a question about who owned what. Right now it's 17 "I" to 16 "we", and your answer could shift that a lot toward "I".
In the draft I call Dana "our PM", since colleagues' names usually stay off a public page. Tell me if you'd rather name her.
The title says what changed rather than how much, since there's no number yet: "Replacing Routewise's setup wizard with a checklist admins could share with their team". Once we have a metric, it should go in the title.
Platform is left as a question in the front matter, because you didn't mention whether it was a web console.

Scorecard

Section	Status	What's missing
Summary	Thin	No result number for the TL;DR
Context	Complete	Could say why it got prioritised in 2022
Role	Thin	Who designed the checklist, invites and importer, versus the PM, researcher and engineers
Research	Complete	Methods, participants, a quote, and the finding that drove the design
Process	Thin	How the checklist was tested, and work-in-progress images
Setbacks	Complete	Two clear setbacks, each with what changed
Tensions	Thin	How sales came around (or didn't) after the comparison
Solution	Complete (no images yet)	Annotated screens of the checklist and importer
Impact	Thin	No before/after numbers, time period or source
Retro	Complete	Two specific "do differently" points

Questions (answer by voice in any order, rough is fine)

You said setup got way faster and implementation calls dropped a lot. Do you remember roughly how long it took a new customer to go live before and after, and how many implementation calls or hours a typical setup needed before and after, over what period? You mentioned rolling the checklist out to new customers first so you could compare. That comparison is probably the best place to look, along with whatever analytics tool you used, launch or QBR decks, the implementation team's call logs, or asking Dana. If exact numbers are gone, would rough wording like "from weeks to days" be accurate?
How did things end with the sales VP? Once the comparison results came in, did sales keep using the demo mode, switch to demoing the checklist, or something else? And when engineering pushed back on teammate invites, who suggested the limited "task assignee" role, and what did you give up from the original idea?
Which parts did you design yourself (the checklist, the invite flow, the CSV importer), and what did Dana, the researcher and the engineers own? This decides where the draft can say "I".
How did you test the checklist before launch? You said the branching wizard "tested badly". Was that usability sessions with admins, and did the checklist go through the same kind of test?
Quick check on the things you hedged: was the wizard fourteen steps, were there six engineers, did you talk to twelve admins, is "I feel like I'm doing someone else's homework" close enough to quote, and do you know which "couple big deals" it helped with (described generally is fine)?

Answering those would cover the biggest gaps. There are a few smaller ones left (why it got prioritised in 2022, a quote from CS, platform, what happened after you left), and I'll come back to them next round.

Shot list so far

Hero: the final setup checklist with one task assigned to a fleet manager.
Research: a simple map of who holds which data (admin, HR, fleet manager) against the old wizard's steps, marking where admins stalled around step four or five.
Process: early checklist sketches or flows next to the old wizard, captioned with what changed.
Solution: the final checklist with an assigned task, annotated.
Solution: the CSV importer showing row-by-row errors, annotated.

Evals

There are seven test cases. Each one ran three times with the skill and once without it, and was graded against 56 checks in total.

Test case (click to see the prompt)	With the skill	Without
Drafts from a messy ramble Promptok I just talked through one of my old projects into my phone and transcribed it, it's in ramble.txt. I want this as a case study for my product design portfolio. can you turn it into a proper draft? Input files: ramble.txt Passes whenA markdown case study draft with all ten sections (summary, context, role, research, process, setbacks, tensions, solution, impact, retro), an outcome-stating title, a 1-2 sentence TL;DR, first-person ownership, and no facts or numbers that are not in the transcript; missing metrics appear as visible placeholders.	27/27	6/9
Asks for missing metrics Prompthere's a transcript of me rambling about the routewise onboarding project (ramble.txt). draft it as a portfolio case study and then tell me what you still need from me to make it really strong Input files: ramble.txt Passes whenA draft plus a short scorecard of section strength and 3-5 targeted follow-up questions, led by impact metrics (baseline, result, timeframe, measurement source), plus questions on tensions or setbacks, tailored to the project.	20/21	4/7
Folds in follow-up answers PromptI answered your questions, my answers are in answers.txt. update the case study draft at case-studies/routewise-onboarding.md with them Input files: answers.txt, case-studies/routewise-onboarding.md, case-studies/routewise-onboarding.source.md Passes whenThe draft is updated in place: the impact section now states 38 to 11 days contract-to-go-live (Salesforce, first two quarters after launch) and roughly 9 to 3 implementation calls per customer; the tensions section ends with the sales VP becoming a fan; the unknown deals figure is not invented; remaining gaps are listed.	22/24	6/8
Refuses to invent from thin input Promptwrite a case study for my portfolio: I redesigned the search page at Kitewell, a travel booking startup, in 2020. It made search way better and users liked it. Passes whenBecause the input is thin, the response does not fabricate research, personas, quotes, metrics or stakeholder conflicts; it produces at most a skeleton with clear placeholders and asks targeted questions to fill it.	21/21	5/7
Respects an NDA Promptturn this into a portfolio case study, it's from my agency days (ramble.txt) Input files: ramble.txt Passes whenA case study draft that anonymises the NDA client (no 'Halvorsen Mutual'), flags nda: true or equivalent, advises redacting or recreating screens, and still covers all sections from what was said, with the handle-time result left as a placeholder.	21/21	6/7
Fits an unfamiliar page template PromptMy Routewise case study is finished (routewise-onboarding.md). Put it on my portfolio site in site/, it should look like my other project pages. Input files: routewise-onboarding.md, site/index.html, site/projects/ledgerly.html, site/site.css Passes whenA new site/projects/routewise-onboarding.html built from the Ledgerly page's markup (cs-hero, cs-title, cs-meta, cs-block sections with Overview/Challenge/Approach/Results/Learnings or similar), with all case-study content mapped into those slots (setbacks and tensions kept, e.g. under Challenge or an added block), the index card linked to it, the mapping explained, and a case-study.config.json saved.	30/30	5/10
Builds a page when there is none PromptPut my finished Routewise case study (routewise-onboarding.md) on my site in site/. I don't have a project page yet. Input files: routewise-onboarding.md, site/index.html, site/style.css Passes whenA new case study page is scaffolded in the site's existing style (links style.css, same fonts/tone), with a container per section (data-section attributes or equivalent), linked from the Work list in index.html, and a case-study.config.json written.	24/24	5/8
Total	165/168, 98%	37/56, 66%

"Without" means Claude with no skill installed, graded against the same checks. With the skill, each run takes about twice as long, 97 seconds on average compared with 43, because it reads its reference files and runs its checker. Results varied by about 5 points from run to run. The three failed checks were a subtraction mistake, a time period attached to the wrong number, and one question that didn't ask how a number was measured. The skill was updated to fix each one. These results are from the version before the writing style rules were added.

A closer look at one test: fitting an unfamiliar template

In this test, the skill puts a finished case study on a portfolio whose project pages have five sections: Overview, Challenge, Approach, Results and Learnings. It has to fit ten sections into those five.

Check	With the skill	Without
A new page for the Routewise case study exists under site/projects/.	Pass	Pass
The new page reuses the Ledgerly page's structure and classes (cs-hero, cs-title, cs-meta, cs-block) rather than inventing a new layout or stylesheet.	Pass	Pass
The content from the draft's setbacks and tensions sections (the branching wizard that tested badly, the sales VP objection) appears somewhere on the new page.	Pass	Pass
The impact content and the retro/lessons content both appear on the new page, in slots like Results and Learnings.	Pass	Fail
The 'Routewise onboarding' card in site/index.html now links to the new page.	Pass	Pass
A case-study.config.json (or similarly named mapping file) records which draft sections map to which page slots.	Pass	Fail
The response explains the section-to-slot mapping to the user, including any sections that were merged.	Pass	Fail
The new case study page shows no raw [[...]] placeholder text to visitors; open gaps are hidden (e.g. as HTML comments) or otherwise not rendered.	Pass	Pass
The new page does not contain <img> elements pointing at image files that do not exist in site/.	Pass	Fail
Because the draft still has open metric/ask placeholders, the home page keeps a 'coming soon' marker for this project (or the reply explicitly asks the user before removing it), rather than silently presenting it as finished.	Pass	Fail

Does it start when it should?

A skill only helps if Claude uses it. Each request below ran three times in a clean project with the skill installed. The ones marked "should not" use similar words but ask for something else.

Request	Expected	Started the skill
ok so I want to write up the Apttus CPQ redesign for my portfolio. I led the quoting flow redesign 2019-2020, worked with like 4 PMs, we killed the old approval matrix... let me just keep talking and you turn it into something good	Should start	3/3
can you help me write a case study about the design system I built at Axcient? I have a bunch of notes in notes/axcient.txt but they're a mess	Should start	3/3
I have a portfolio review with a hiring manager at Figma next week and my Stash Management project page is just screenshots. help me turn it into a real story with problem, process and results	Should start	3/3
here's a transcript of me talking about the agent experiences studio project, I recorded it on a walk. make it portfolio-ready and tell me what's missing	Should start	3/3
what's my Lightning Design System write-up missing? recruiters keep saying my case studies don't show impact. the draft is at case-studies/slds.md	Should start	3/3
I'm a senior product designer applying for staff roles. I need to document the checkout redesign I did at Kitewell in a way that shows how I handled the fight with the marketing team over the upsell modal	Should start	3/3
turn these bullet points into a UX case study: - redesigned claims intake for insurance agents - shadowed call center - single screen - handle time down	Should start	3/3
my case study for the routewise onboarding project is done, can you put it on my site so it matches the other project pages in /projects	Should start	3/3
I keep forgetting what to include in a product design case study. interview me about my last project at Brightwheel and write the case study from my answers	Should start	3/3
help me write the project page for my portfolio about the mobile banking app I designed at Chime, I want it to read like the ones on workbymark.nl	Should start	3/3
create a portfolio piece for the Stash Management app I designed and built as a founder, I'll tell you about it	Should start	3/3
I need a project write-up for my site about the Salesforce agent experiences work. can you help me put it together?	Should start	3/3
can you make a portfolio entry for my work on the Prime Nevada site? it was a small brand + web project	Should start	3/3
recruiter asked for 2 work samples from my product design work. help me write up the Apttus quoting redesign as one	Should start	3/3
I have a PM interview at Stripe tomorrow and they'll give me a product case question like 'how would you improve Stripe Checkout'. can you run me through a mock case?	Should not	0/3
write a customer success story for our website about how Acme Logistics cut delivery costs 18% using our routing software, quote their COO	Should not	0/3
summarize the findings from this usability study report (research/q3-usability.pdf) into 5 key insights for the team	Should not	0/3
the case study cards on my portfolio homepage look cramped on mobile, can you fix the grid and spacing in styles.css	Should not	0/3
rewrite these three resume bullets about my Salesforce design system work so they're punchier and fit on one line each	Should not	0/3
help me plan user research for a new onboarding flow: who to recruit, interview guide, and how many sessions	Should not	0/3
I'm in an MBA class and need to analyze the Harvard Business School case on Netflix's pivot to streaming, what are the key strategic decisions?	Should not	0/3
write documentation for the Button component in our design system: usage guidelines, do's and don'ts, accessibility notes	Should not	0/3
draft a LinkedIn post announcing that the new onboarding checklist we shipped at Routewise is live, upbeat but not cringey	Should not	0/3
add 'coming soon' badges to the three unfinished project cards on my portfolio index.html	Should not	0/3
design a new layout for the portfolio case study page in index.html, bigger hero images and a sticky table of contents	Should not	0/3
review my portfolio site's performance and accessibility, the case study images load slowly	Should not	0/3

All 14 requests that should start the skill did, every time. None of the other 12 did. This was re-run after the writing style changes.

Reference

What's in the package

`SKILL.md`	The main instructions: capture, draft, score and ask questions, preview, and add to your page
`references/structure.md`	The ten sections, with examples
`references/rubric.md`	The checklist hiring managers use, with sources
`references/question-bank.md`	Follow-up questions for each section, including questions about numbers
`references/apply.md`	How it finds your page or doc and fits the case study into it
`templates/case-study.md`	The markdown skeleton
`scripts/check_draft.py`	Checks for missing sections, numbers that don't match your notes, open placeholders, em dashes and arrows

Sources

If you have questions or ideas, email me at alanweibel@gmail.com.