Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Write a help-center / knowledge-base article that actually resolves the issue and deflects tickets. Use when asked to write a help doc, KB article, FAQ entry, how-to, or support documentation. Produces a findable, skimmable article — task-based title, the answer up front, numbered steps, screenshots-to-add markers, troubleshooting, and related links — written so users self-serve instead of contacting support.
.claude/skills/mohitagw15856-help-center-article/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 48% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 36% | 0% |
A help article's job is deflection: the user finds it, solves their problem, and never opens a ticket. That requires a findable title (what they'd search), the answer up front (not after three paragraphs of preamble), and skimmable steps. This skill writes that — task-based, scannable, and SEO/search-friendly so it surfaces both in your help center and in Google.
Ask for these only if they aren't already provided:
e.g. "How to reset your password" / "Why is my export failing?" — not "Password Management."
1. Short answer (TL;DR) — resolve it in 1–2 sentences right at the top for the people who just need the quick fix. Then the detail for those who need it.
2. Before you start — prerequisites/permissions, if any (so step 3 doesn't fail silently).
3. Steps — numbered, one action per step, in the user's language. Mark where a screenshot] should go. Bold the buttons/menu names they'll click.
4. Troubleshooting — the 2–4 common "it didn't work" cases and the fix for each. This is what prevents the follow-up ticket.
5. Related articles — links to the adjacent tasks (the next thing they'll need).
SEO/findability note: use the words users actually type (synonyms in the body), keep the title a real question/task, and front-load the answer.
Knowledge-base / technical-writing practice — task-based titles, answer-first, scannable steps, search-optimised, ticket-deflection focus.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 9,068 | 6,842 | -25% | 1 | 1 | 0% | 1,614 | 2,211 | +37% | 0 | 0 | — |
case-02 | fail→pass | 14,132 | 11,022 | -22% | 1 | 1 | 0% | 2,607 | 2,876 | +10% | 0 | 0 | — |
case-03 | fail→pass | 12,089 | 8,941 | -26% | 1 | 1 | 0% | 2,187 | 2,363 | +8% | 0 | 0 | — |
case-04 | pass→pass | 12,488 | 10,157 | -19% | 1 | 1 | 0% | 2,051 | 2,507 | +22% | 0 | 0 | — |
case-05 | pass→pass | 12,677 | 10,485 | -17% | 1 | 1 | 0% | 2,319 | 2,693 | +16% | 0 | 0 | — |
case-06 | fail→pass | 9,631 | 10,580 | +10% | 1 | 1 | 0% | 1,711 | 2,535 | +48% | 0 | 0 | — |
case-07 | fail→fail | 8,947 | 8,380 | -6% | 1 | 1 | 0% | 1,744 | 2,315 | +33% | 0 | 0 | — |
case-08 | fail→pass | 10,671 | 8,749 | -18% | 1 | 1 | 0% | 1,802 | 2,446 | +36% | 0 | 0 | — |
case-09 | fail→pass | 8,920 | 8,756 | -2% | 1 | 1 | 0% | 1,677 | 2,329 | +39% | 0 | 0 | — |
case-10 | pass→pass | 10,801 | 11,388 | +5% | 1 | 1 | 0% | 2,048 | 2,702 | +32% | 0 | 0 | — |
case-11 | fail→pass | 11,587 | 6,242 | -46% | 1 | 1 | 0% | 2,017 | 1,844 | -9% | 0 | 0 | — |
case-12 | fail→pass | 10,654 | 11,795 | +11% | 1 | 1 | 0% | 1,953 | 2,663 | +36% | 0 | 0 | — |
case-13 | pass→pass | 12,132 | 9,685 | -20% | 1 | 1 | 0% | 2,234 | 2,492 | +12% | 0 | 0 | — |
case-14 | pass→pass | 17,533 | 10,992 | -37% | 1 | 1 | 0% | 3,509 | 2,607 | -26% | 0 | 0 | — |
case-15 | fail→pass | 10,937 | 9,054 | -17% | 1 | 1 | 0% | 1,906 | 2,400 | +26% | 0 | 0 | — |
case-16 | pass→pass | 13,275 | 8,967 | -32% | 1 | 1 | 0% | 2,282 | 2,275 | -0% | 0 | 0 | — |
case-17 | pass→pass | 11,180 | 8,360 | -25% | 1 | 1 | 0% | 2,018 | 1,987 | -2% | 0 | 0 | — |
case-18 | pass→pass | 11,167 | 7,861 | -30% | 1 | 1 | 0% | 1,915 | 2,171 | +13% | 0 | 0 | — |
case-19 | pass→pass | 12,743 | 7,038 | -45% | 1 | 1 | 0% | 2,496 | 2,039 | -18% | 0 | 0 | — |
case-20 | pass→pass | 14,999 | 13,865 | -8% | 1 | 1 | 0% | 2,409 | 2,860 | +19% | 0 | 0 | — |
case-21 | pass→pass | 18,000 | 16,411 | -9% | 1 | 1 | 0% | 2,884 | 3,516 | +22% | 0 | 0 | — |
case-22 | pass→pass | 16,316 | 12,791 | -22% | 1 | 1 | 0% | 3,713 | 3,235 | -13% | 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 +41 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.