Install any skill in seconds. Free to start, no credit card required.
Get Started Free →When the user needs to write, review, or improve a job posting for a startup role.
.claude/skills/mkurman-job-description/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 67% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 76% | 0% |
Activate when the user asks to create a new job posting, rewrite an existing one, or get feedback on a draft JD. Also activate when the user is preparing to hire for a new role and needs to define it clearly before sourcing candidates.
A complete, ready-to-post job description in markdown with the following sections:
interview-kit — After posting the JD, design the interview loop and scorecards for this role.sourcing-outreach — Use the JD as the basis for personalized outreach messages to passive candidates.employer-brand — Ensure the JD tone and content align with your careers page and employer brand narrative.Prompt: "We need to hire our first product designer. We're a 12-person B2B SaaS startup, Series A, remote-first."
Good output snippet:
# Product Designer — Remote (US/Canada time zones)
Join Acme as our founding designer and shape the product experience
from the ground up. You won't be handed a pixel-perfect spec — you'll
partner directly with engineering and customers to figure out what to
build and how it should feel.
## What You'll Do
- Own the end-to-end design process for our core B2B platform
- Run discovery calls with customers weekly to ground designs in real workflows
- Build and maintain our design system as the foundation for a future team
- Collaborate async-first with engineering using Figma and Loom
...Prompt: "Review this JD for red flags."
Good output snippet:
Issues found:
1. "Must have 10+ years in Python" — This filters out strong mid-career
candidates. Consider "deep fluency in Python" instead.
2. "Fast-paced environment" — This is vague. Replace with something
concrete: "We ship weekly and iterate based on customer feedback."
3. No salary range listed — Adding a range will increase your applicant
pool significantly.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 17,140 | 13,155 | -23% | 1 | 1 | 0% | 2,704 | 3,559 | +32% | 0 | 0 | — |
case-02 | fail→fail | 16,147 | 15,192 | -6% | 1 | 1 | 0% | 2,478 | 3,855 | +56% | 0 | 0 | — |
case-03 | pass→pass | 21,530 | 15,686 | -27% | 1 | 1 | 0% | 3,328 | 3,814 | +15% | 0 | 0 | — |
case-04 | pass→pass | 8,892 | 10,169 | +14% | 1 | 1 | 0% | 1,247 | 2,869 | +130% | 0 | 0 | — |
case-05 | pass→pass | 15,381 | 12,274 | -20% | 1 | 1 | 0% | 2,106 | 2,988 | +42% | 0 | 0 | — |
case-06 | pass→pass | 12,412 | 11,484 | -7% | 1 | 1 | 0% | 2,106 | 3,286 | +56% | 0 | 0 | — |
case-07 | fail→pass | 14,737 | 12,489 | -15% | 1 | 1 | 0% | 2,170 | 3,268 | +51% | 0 | 0 | — |
case-08 | fail→pass | 11,876 | 14,146 | +19% | 1 | 1 | 0% | 2,006 | 3,586 | +79% | 0 | 0 | — |
case-09 | fail→fail | 11,484 | 12,069 | +5% | 1 | 1 | 0% | 1,878 | 3,218 | +71% | 0 | 0 | — |
case-10 | fail→pass | 13,515 | 12,842 | -5% | 1 | 1 | 0% | 2,027 | 3,376 | +67% | 0 | 0 | — |
case-11 | fail→pass | 11,089 | 11,274 | +2% | 1 | 1 | 0% | 1,827 | 3,217 | +76% | 0 | 0 | — |
case-12 | fail→fail | 14,188 | 9,271 | -35% | 1 | 1 | 0% | 2,093 | 2,677 | +28% | 0 | 0 | — |
case-13 | fail→fail | 10,319 | 10,776 | +4% | 1 | 1 | 0% | 1,706 | 3,104 | +82% | 0 | 0 | — |
case-14 | pass→pass | 13,963 | 17,000 | +22% | 1 | 1 | 0% | 2,109 | 4,015 | +90% | 0 | 0 | — |
case-15 | pass→pass | 14,699 | 14,625 | -1% | 1 | 1 | 0% | 2,264 | 3,555 | +57% | 0 | 0 | — |
case-16 | pass→pass | 10,945 | 11,054 | +1% | 1 | 1 | 0% | 1,789 | 3,207 | +79% | 0 | 0 | — |
case-17 | pass→pass | 11,545 | 13,367 | +16% | 1 | 1 | 0% | 1,704 | 3,576 | +110% | 0 | 0 | — |
case-18 | fail→pass | 12,850 | 13,900 | +8% | 1 | 1 | 0% | 2,048 | 3,536 | +73% | 0 | 0 | — |
case-19 | pass→pass | 12,377 | 13,684 | +11% | 1 | 1 | 0% | 1,896 | 3,431 | +81% | 0 | 0 | — |
case-20 | pass→pass | 14,888 | 15,332 | +3% | 1 | 1 | 0% | 2,112 | 3,686 | +75% | 0 | 0 | — |
case-21 | pass→pass | 8,802 | 7,812 | -11% | 1 | 1 | 0% | 1,274 | 2,490 | +95% | 0 | 0 | — |
case-22 | pass→pass | 13,860 | 14,033 | +1% | 1 | 1 | 0% | 2,109 | 3,610 | +71% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +27 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.