Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Write a persuasive prior-authorization / medical-necessity letter to an insurer. Use when asked to write a prior authorization letter, a letter of medical necessity, or to appeal a denied treatment/medication/procedure. Produces a structured letter — patient and request, clinical justification tied to guidelines, treatments tried, and the specific approval asked for — ready for clinician review and signature.
.claude/skills/mohitagw15856-prior-authorization-letter/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 84% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 23% | 0% |
A prior-auth or medical-necessity letter succeeds when it connects this patient's clinical facts to the insurer's coverage criteria — clearly, with evidence, and with the exact request spelled out. This skill structures that argument so the reviewer can approve it quickly, and so an appeal addresses the stated denial reason head-on.
> Clinical-safety note: this is a documentation aid, not medical advice. The clinical justification must > reflect the treating clinician's judgement and the patient's actual record; the clinician must review, verify, > and sign before submission. Do not invent diagnoses, codes, history, or evidence.
Given the treatment and a diagnosis, produce the full letter anyway — structure the argument and insert the standard elements, marking patient-specific facts (codes, dates, prior treatments) to be confirmed rather than inventing them. For an appeal, infer and directly rebut the likely denial reason if it's stated. Never fabricate clinical history or citations.
Ask for these only if they aren't already provided (else mark to confirm):
guidelines/evidence and the insurer's likely coverage criteria.
For an appeal, add a section that quotes the denial reason and rebuts it specifically.
Close with a list of facts to confirm before sending and a clinician-sign-off reminder.
Utilization-management correspondence practice — medical-necessity argumentation tied to coverage criteria, step-therapy documentation, and targeted appeals.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 18,211 | 21,484 | +18% | 1 | 1 | 0% | 3,548 | 4,110 | +16% | 0 | 0 | — |
case-02 | pass→pass | 19,532 | 14,492 | -26% | 1 | 1 | 0% | 2,636 | 3,455 | +31% | 0 | 0 | — |
case-03 | pass→pass | 14,497 | 30,987 | +114% | 1 | 1 | 0% | 2,750 | 4,601 | +67% | 0 | 0 | — |
case-04 | pass→pass | 16,576 | 16,935 | +2% | 1 | 1 | 0% | 1,957 | 2,770 | +42% | 0 | 0 | — |
case-05 | fail→pass | 18,997 | 23,496 | +24% | 1 | 1 | 0% | 2,609 | 3,730 | +43% | 0 | 0 | — |
case-06 | pass→pass | 21,908 | 22,331 | +2% | 1 | 1 | 0% | 2,403 | 3,339 | +39% | 0 | 0 | — |
case-07 | pass→pass | 19,104 | 20,069 | +5% | 1 | 1 | 0% | 2,388 | 3,541 | +48% | 0 | 0 | — |
case-08 | pass→pass | 15,167 | 21,896 | +44% | 1 | 1 | 0% | 2,167 | 4,288 | +98% | 0 | 0 | — |
case-09 | pass→pass | 17,061 | 17,741 | +4% | 1 | 1 | 0% | 1,890 | 3,029 | +60% | 0 | 0 | — |
case-10 | fail→pass | 17,717 | 21,207 | +20% | 1 | 1 | 0% | 2,015 | 3,716 | +84% | 0 | 0 | — |
case-11 | pass→pass | 18,328 | 22,420 | +22% | 1 | 1 | 0% | 2,402 | 4,001 | +67% | 0 | 0 | — |
case-12 | fail→pass | 22,424 | 20,692 | -8% | 1 | 1 | 0% | 2,719 | 3,500 | +29% | 0 | 0 | — |
case-13 | fail→pass | 21,128 | 21,526 | +2% | 1 | 1 | 0% | 2,522 | 3,547 | +41% | 0 | 0 | — |
case-14 | fail→pass | 23,142 | 17,967 | -22% | 1 | 1 | 0% | 2,445 | 3,013 | +23% | 0 | 0 | — |
case-15 | fail→pass | 18,821 | 19,703 | +5% | 1 | 1 | 0% | 2,733 | 3,686 | +35% | 0 | 0 | — |
case-16 | pass→pass | 18,228 | 22,009 | +21% | 1 | 1 | 0% | 2,260 | 3,575 | +58% | 0 | 0 | — |
case-17 | pass→pass | 18,530 | 23,498 | +27% | 1 | 1 | 0% | 2,645 | 3,470 | +31% | 0 | 0 | — |
case-18 | pass→pass | 16,444 | 23,019 | +40% | 1 | 1 | 0% | 1,781 | 3,216 | +81% | 0 | 0 | — |
case-19 | pass→pass | 23,788 | 19,940 | -16% | 1 | 1 | 0% | 2,556 | 3,698 | +45% | 0 | 0 | — |
case-20 | fail→pass | 16,747 | 22,183 | +32% | 1 | 1 | 0% | 2,610 | 3,577 | +37% | 0 | 0 | — |
case-21 | pass→pass | 15,310 | 24,521 | +60% | 1 | 1 | 0% | 2,644 | 4,042 | +53% | 0 | 0 | — |
case-22 | fail→pass | 23,577 | 14,765 | -37% | 1 | 1 | 0% | 3,215 | 3,626 | +13% | 0 | 0 | — |
case-23 | pass→pass | 19,738 | 23,671 | +20% | 1 | 1 | 0% | 2,399 | 3,661 | +53% | 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. 23 cases were attempted. The headline lift of +35 percentage points is the difference between those two pass rates over the 23 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.