Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user wants to tailor a workflow for a specific industry, domain, or vertical with specialized expertise, terminology, and guardrails.
.claude/skills/sharpdeveye-specialize/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 99% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 41% | 0% |
Invoke /agent-workflow — it contains workflow principles, anti-patterns, and the Context Gathering Protocol. Follow the protocol before proceeding — if no workflow context exists yet, you MUST run /teach-maestro first.
Transform a general-purpose workflow into a domain expert.
markdown## Generic: You are an assistant that analyzes documents. ## Specialized (legal): You are a senior legal analyst specializing in contract review. You understand common law jurisdictions, standard contract clauses, and the difference between representations and warranties. Always caveat that this is not legal advice.
| Domain | Evaluation Criteria | |--------|-------------------| | Legal | Clause completeness, regulatory compliance, jurisdiction accuracy | | Medical | Clinical accuracy, guideline adherence, contraindication checks | | Financial | Calculation accuracy, regulatory disclosure, risk assessment | | Code | Test coverage, security vulnerabilities, performance | | Customer Support | Tone, escalation accuracy, resolution completeness |
After specialization, run /evaluate with domain-specific scenarios, then /guard to add domain-appropriate safety guardrails.
NEVER:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | fail→pass | 11,872 | 4,944 | -58% | 1 | 1 | 0% | 1,842 | 1,148 | -38% | 0 | 0 | — |
case-01 | fail→pass | 15,516 | 16,244 | +5% | 1 | 1 | 0% | 2,388 | 2,946 | +23% | 0 | 0 | — |
case-02 | fail→pass | 20,521 | 36,843 | +80% | 1 | 1 | 0% | 3,331 | 6,626 | +99% | 0 | 0 | — |
case-03 | fail→fail | 18,900 | 24,275 | +28% | 1 | 1 | 0% | 2,759 | 4,073 | +48% | 0 | 0 | — |
case-04 | pass→pass | 9,105 | 8,051 | -12% | 1 | 1 | 0% | 1,374 | 1,754 | +28% | 0 | 0 | — |
case-05 | pass→pass | 10,281 | 8,840 | -14% | 1 | 1 | 0% | 1,567 | 1,786 | +14% | 0 | 0 | — |
case-06 | pass→pass | 10,770 | 8,741 | -19% | 1 | 1 | 0% | 1,648 | 1,833 | +11% | 0 | 0 | — |
case-07 | pass→pass | 12,148 | 4,689 | -61% | 1 | 1 | 0% | 1,896 | 1,036 | -45% | 0 | 0 | — |
case-08 | pass→pass | 11,113 | 10,536 | -5% | 1 | 1 | 0% | 1,702 | 1,973 | +16% | 0 | 0 | — |
case-09 | pass→pass | 8,802 | 4,113 | -53% | 1 | 1 | 0% | 1,328 | 1,118 | -16% | 0 | 0 | — |
case-10 | pass→pass | 16,887 | 13,649 | -19% | 1 | 1 | 0% | 2,537 | 2,625 | +3% | 0 | 0 | — |
case-11 | fail→pass | 9,805 | 4,974 | -49% | 1 | 1 | 0% | 1,449 | 1,139 | -21% | 0 | 0 | — |
case-13 | pass→pass | 15,409 | 7,510 | -51% | 1 | 1 | 0% | 2,237 | 1,527 | -32% | 0 | 0 | — |
case-14 | pass→pass | 13,923 | 14,323 | +3% | 1 | 1 | 0% | 1,976 | 2,569 | +30% | 0 | 0 | — |
case-15 | pass→pass | 6,764 | 4,978 | -26% | 1 | 1 | 0% | 1,115 | 1,153 | +3% | 0 | 0 | — |
case-16 | pass→pass | 6,471 | 2,847 | -56% | 1 | 1 | 0% | 890 | 848 | -5% | 0 | 0 | — |
case-17 | pass→pass | 13,988 | 13,182 | -6% | 1 | 1 | 0% | 2,155 | 2,442 | +13% | 0 | 0 | — |
case-18 | pass→pass | 15,781 | 14,881 | -6% | 1 | 1 | 0% | 2,044 | 2,619 | +28% | 0 | 0 | — |
case-19 | pass→pass | 12,665 | 8,066 | -36% | 1 | 1 | 0% | 1,730 | 1,548 | -11% | 0 | 0 | — |
case-20 | pass→pass | 13,693 | 37,386 | +173% | 1 | 1 | 0% | 2,694 | 3,930 | +46% | 0 | 0 | — |
case-21 | fail→pass | 9,378 | 9,444 | +1% | 1 | 1 | 0% | 1,466 | 2,066 | +41% | 0 | 0 | — |
case-22 | fail→fail | 1,831 | 11,299 | +517% | 1 | 1 | 0% | 253 | 2,347 | +828% | 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 +23 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.