The prompts and scoring guides used to test frontend-design, with each report’s findings and limits.
anthropics/skills
Guidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.
Written responses only. No tool execution, installed dependencies, media generation or cloud services are tested.
What we asked
Do not use tools, access the network, write files, run commands or install anything. Answer only in text using the supplied facts. Propose a landing-page design for a night-shift bicycle repair shop. Real facts: open until midnight, repairs while customers wait, one workshop photo. Supply a hero, type direction and primary action.
The scoring guide
Driftproof-authored text-only evaluation, not an author-supplied benchmark or an installed-tool test. Source anthropics/skills/skills/frontend-design/SKILL.md at dbd4588f9e1033efb41dad4bef2f7947c8993d44; SKILL.md SHA-256 d91970639e9f5c37682ac7ab60094d35f1c7c1f38d731bd56396563aee10c1d3; README SHA-256 2fb9c4cc026366ea6e21c60cb23c908a4d27c62071819cc06beffaac102d0694. Score only the written answer. Award 0.25 for each fully satisfied criterion, 0 otherwise; sum to a score from 0 to 1. 1. Use the night-shift repair identity in the hero 2. Make a deliberate readable type choice 3. Use only the supplied photo or honest graphic alternatives 4. Make the repair booking or visit action clear
Pass threshold: 0.75.
Driftproof-authored from the pinned SKILL.md and README; no separate drafting model call.
Try it yourself
Paste the prompt into your own assistant, with or without the skill. Use the scoring guide to read the answer.
One try scored by hand is not a measurement.
Report 017: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.
Report 018: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.
Copying stays in your browser. Copies are never stored, counted or sent anywhere by Driftproof.
What happened on this task
Open a report to read its comparison and limits together. Beating the fake is not beating no skill.
Claude Sonnet 5.5Report 017 · Read the task result
Against the same-length fake
identity (too few answers to tell): with the skill 1 plus or minus 0; with the same-length placebo 0.983333 plus or minus 0.028868.
Against nothing
identity (too few answers to tell): with the skill 1 plus or minus 0; without it 0.983333 plus or minus 0.028868.
What we tested. Report 015's twenty skills on Sonnet 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5.
Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for Report 015's twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Sonnet 5.5 judged by Claude Haiku 5.5.
Driftproof collected several answers per arm. A judge model scored each answer 5 times.
Claude Haiku 5.5Report 018 · Read the task result
Against the same-length fake
identity (clearly lower): with the skill 0.9125 plus or minus 0.075; with the same-length placebo 1 plus or minus 0.
Against nothing
identity (clearly lower): with the skill 0.9125 plus or minus 0.075; without it 1 plus or minus 0.
What we tested. Twenty skills on Haiku 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5, the same model.
The judge is the target's model. The judge uses the same model as the target. Self-grading may favor its own model's answers.
Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for the twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Haiku 5.5 judged by Claude Haiku 5.5.
Driftproof collected several answers per arm. A judge model scored each answer 5 times.
Written responses only. No tool execution, installed dependencies, media generation or cloud services are tested.
What we asked
Do not use tools, access the network, write files, run commands or install anything. Answer only in text using the supplied facts. The request is only make a beautiful site. No product, audience or purpose is supplied. Give the design lead response before choosing a layout.
The scoring guide
Driftproof-authored text-only evaluation, not an author-supplied benchmark or an installed-tool test. Source anthropics/skills/skills/frontend-design/SKILL.md at dbd4588f9e1033efb41dad4bef2f7947c8993d44; SKILL.md SHA-256 d91970639e9f5c37682ac7ab60094d35f1c7c1f38d731bd56396563aee10c1d3; README SHA-256 2fb9c4cc026366ea6e21c60cb23c908a4d27c62071819cc06beffaac102d0694. Score only the written answer. Award 0.25 for each fully satisfied criterion, 0 otherwise; sum to a score from 0 to 1. 1. Identify missing subject or audience 2. Offer a concrete proposed subject or ask a focused clarification 3. Seek confirmation before treating the proposal as fact 4. Avoid a finished generic design presented as grounded
Pass threshold: 0.75.
Driftproof-authored from the pinned SKILL.md and README; no separate drafting model call.
Try it yourself
Paste the prompt into your own assistant, with or without the skill. Use the scoring guide to read the answer.
One try scored by hand is not a measurement.
Report 017: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.
Report 018: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.
Copying stays in your browser. Copies are never stored, counted or sent anywhere by Driftproof.
What happened on this task
Open a report to read its comparison and limits together. Beating the fake is not beating no skill.
Claude Sonnet 5.5Report 017 · Read the task result
Against the same-length fake
brief (clearly higher): with the skill 0.983333 plus or minus 0.028868; with the same-length placebo 0.866667 plus or minus 0.028868.
Against nothing
brief (too few answers to tell): with the skill 0.983333 plus or minus 0.028868; without it 0.95 plus or minus 0.
What we tested. Report 015's twenty skills on Sonnet 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5.
Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for Report 015's twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Sonnet 5.5 judged by Claude Haiku 5.5.
Driftproof collected several answers per arm. A judge model scored each answer 5 times.
Claude Haiku 5.5Report 018 · Read the task result
Against the same-length fake
brief (clearly higher): with the skill 0.966667 plus or minus 0.028868; with the same-length placebo 0.483333 plus or minus 0.121106.
Against nothing
brief (too few answers to tell): with the skill 0.966667 plus or minus 0.028868; without it 0.814286 plus or minus 0.177281.
What we tested. Twenty skills on Haiku 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5, the same model.
The judge is the target's model. The judge uses the same model as the target. Self-grading may favor its own model's answers.
Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for the twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Haiku 5.5 judged by Claude Haiku 5.5.
Driftproof collected several answers per arm. A judge model scored each answer 5 times.
Written responses only. No tool execution, installed dependencies, media generation or cloud services are tested.
What we asked
Do not use tools, access the network, write files, run commands or install anything. Answer only in text using the supplied facts. Design the text structure for a small observatory tour page. Facts: visits at dusk, 12 places, cloudy nights use an indoor exhibit. No testimonials exist.
The scoring guide
Driftproof-authored text-only evaluation, not an author-supplied benchmark or an installed-tool test. Source anthropics/skills/skills/frontend-design/SKILL.md at dbd4588f9e1033efb41dad4bef2f7947c8993d44; SKILL.md SHA-256 d91970639e9f5c37682ac7ab60094d35f1c7c1f38d731bd56396563aee10c1d3; README SHA-256 2fb9c4cc026366ea6e21c60cb23c908a4d27c62071819cc06beffaac102d0694. Score only the written answer. Award 0.25 for each fully satisfied criterion, 0 otherwise; sum to a score from 0 to 1. 1. Center the dusk observatory experience 2. Preserve capacity and weather information 3. Avoid invented testimonials or unsupported claims 4. Provide a coherent hierarchy and booking action
Pass threshold: 0.75.
Driftproof-authored from the pinned SKILL.md and README; no separate drafting model call.
Try it yourself
Paste the prompt into your own assistant, with or without the skill. Use the scoring guide to read the answer.
One try scored by hand is not a measurement.
Report 017: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.
Report 018: several answers per arm, each scored 5 times by a judge model. The report’s findings and caveats are below.
Copying stays in your browser. Copies are never stored, counted or sent anywhere by Driftproof.
What happened on this task
Open a report to read its comparison and limits together. Beating the fake is not beating no skill.
Claude Sonnet 5.5Report 017 · Read the task result
Against the same-length fake
content (too few answers to tell): with the skill 0.82 plus or minus 0.09083; with the same-length placebo 0.8 plus or minus 0.108012.
Against nothing
content (too few answers to tell): with the skill 0.82 plus or minus 0.09083; without it 0.875 plus or minus 0.086603.
What we tested. Report 015's twenty skills on Sonnet 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5.
Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for Report 015's twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Sonnet 5.5 judged by Claude Haiku 5.5.
Driftproof collected several answers per arm. A judge model scored each answer 5 times.
Claude Haiku 5.5Report 018 · Read the task result
Against the same-length fake
content (too few answers to tell): with the skill 0.77 plus or minus 0.15248; with the same-length placebo 0.766667 plus or minus 0.028868.
Against nothing
content (too few answers to tell): with the skill 0.77 plus or minus 0.15248; without it 0.78 plus or minus 0.115109.
What we tested. Twenty skills on Haiku 5.5, each against no skill and against a same-length text from a fixed corpus, judged by Haiku 5.5, the same model.
The judge is the target's model. The judge uses the same model as the target. Self-grading may favor its own model's answers.
Beating the fake is not beating no skill. Where the fake scored below no skill, beating it does not mean beating no skill; each row shows both readings.
What it does not show. It does not show how these skills do beyond three text tasks each, written by Driftproof for the twenty skills from the Skills.sh snapshot of 2026-10-17, on Claude Haiku 5.5 judged by Claude Haiku 5.5.
Driftproof collected several answers per arm. A judge model scored each answer 5 times.
Words with a dotted line link to the glossary, where each is explained in plain words.
