Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Write a credible, specific letter of recommendation or reference. Use when asked to write a reference letter, a letter of recommendation, a character reference, or to recommend someone for a job, school, or tenancy. Produces a structured reference — your relationship, specific evidence of their strengths, a comparative endorsement, and a clear recommendation — tailored to what the reader is deciding.
.claude/skills/mohitagw15856-reference-letter/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 74% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 44% | 0% |
A reference is believed when it's specific: concrete examples beat adjectives, and the reader can tell you actually know the person. This skill writes a letter that establishes your credibility to comment, gives real evidence of the person's strengths, and makes a clear, tailored recommendation for the decision at hand.
Given "write a reference for my report applying for a senior role", write the full letter anyway — infer plausible, concrete examples from the relationship described, clearly marking invented specifics as (example — replace with a real instance) so the writer swaps in true details. Never hand back a hollow template of adjectives.
Ask for these only if they aren't already provided (else infer and label for replacement):
Mark any invented specifics as (example — replace with a real instance). Provide a shorter version if useful.
Recommendation-writing practice — establishing credibility, evidence over adjectives, comparative endorsement, and tailoring to the reader's decision.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 17,353 | 16,814 | -3% | 1 | 1 | 0% | 1,991 | 2,484 | +25% | 0 | 0 | — |
case-02 | fail→pass | 16,502 | 21,199 | +28% | 1 | 1 | 0% | 2,499 | 2,721 | +9% | 0 | 0 | — |
case-03 | pass→pass | 12,050 | 19,723 | +64% | 1 | 1 | 0% | 1,751 | 2,465 | +41% | 0 | 0 | — |
case-04 | fail→pass | 10,716 | 17,562 | +64% | 1 | 1 | 0% | 1,418 | 2,463 | +74% | 0 | 0 | — |
case-05 | pass→pass | 15,424 | 14,149 | -8% | 1 | 1 | 0% | 1,926 | 2,650 | +38% | 0 | 0 | — |
case-06 | fail→pass | 12,960 | 16,276 | +26% | 1 | 1 | 0% | 1,985 | 2,677 | +35% | 0 | 0 | — |
case-07 | fail→pass | 12,772 | 17,170 | +34% | 1 | 1 | 0% | 1,862 | 2,682 | +44% | 0 | 0 | — |
case-08 | fail→pass | 13,366 | 15,033 | +12% | 1 | 1 | 0% | 1,957 | 2,825 | +44% | 0 | 0 | — |
case-09 | pass→pass | 12,346 | 16,963 | +37% | 1 | 1 | 0% | 1,823 | 2,760 | +51% | 0 | 0 | — |
case-10 | fail→pass | 14,759 | 17,418 | +18% | 1 | 1 | 0% | 1,844 | 2,973 | +61% | 0 | 0 | — |
case-11 | fail→pass | 11,951 | 18,075 | +51% | 1 | 1 | 0% | 1,676 | 2,814 | +68% | 0 | 0 | — |
case-12 | fail→pass | 40,903 | 21,110 | -48% | 1 | 1 | 0% | 2,027 | 3,007 | +48% | 0 | 0 | — |
case-13 | fail→pass | 10,815 | 11,623 | +7% | 1 | 1 | 0% | 1,598 | 2,427 | +52% | 0 | 0 | — |
case-14 | fail→pass | 14,036 | 14,241 | +1% | 1 | 1 | 0% | 1,591 | 2,720 | +71% | 0 | 0 | — |
case-15 | fail→pass | 11,486 | 11,982 | +4% | 1 | 1 | 0% | 1,753 | 2,573 | +47% | 0 | 0 | — |
case-16 | pass→pass | 12,017 | 15,336 | +28% | 1 | 1 | 0% | 1,938 | 2,770 | +43% | 0 | 0 | — |
case-17 | fail→pass | 13,401 | 15,116 | +13% | 1 | 1 | 0% | 1,621 | 2,734 | +69% | 0 | 0 | — |
case-18 | fail→pass | 14,499 | 14,064 | -3% | 1 | 1 | 0% | 1,784 | 2,761 | +55% | 0 | 0 | — |
case-19 | pass→pass | 14,642 | 13,584 | -7% | 1 | 1 | 0% | 1,740 | 2,492 | +43% | 0 | 0 | — |
case-20 | pass→pass | 13,824 | 16,978 | +23% | 1 | 1 | 0% | 2,130 | 3,305 | +55% | 0 | 0 | — |
case-21 | pass→pass | 8,431 | 6,043 | -28% | 1 | 1 | 0% | 1,117 | 1,729 | +55% | 0 | 0 | — |
case-22 | pass→pass | 12,402 | 17,554 | +42% | 1 | 1 | 0% | 1,848 | 3,033 | +64% | 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 +59 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.