Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user wants to create templates, extract reusable patterns, document solutions, or build a pattern library from working workflows.
.claude/skills/sharpdeveye-extract-pattern/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 12% | 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.
Turn working solutions into reusable patterns. Every successful workflow contains patterns that are applicable beyond their original context.
Review the workflow and identify components that:
Transform the specific solution into a reusable template:
From specific → To general pattern:
markdown## Pattern: [Name] **Problem**: What recurring problem does this solve? **When to use**: When is this pattern appropriate? **When NOT to use**: When is this pattern inappropriate? **Template**: Copy-pastable starting point with customization markers. **Variants**: Common variations for different contexts. **Pitfalls**: What went wrong during development and how it was fixed. **Examples**: 1-2 concrete examples.
| Workflow Element | Extract As | |-----------------|-----------| | Effective prompt structure | Prompt template with customization points | | Tool chain that works well | Pipeline pattern with data flow diagram | | Error handling strategy | Resilience pattern with implementation guide | | Evaluation approach | Quality assurance pattern with scoring rubric | | Context management technique | Context pattern with budget guidance | | Agent coordination protocol | Orchestration pattern with handoff templates |
After extracting patterns, run /calibrate to ensure the new patterns are consistent with existing project conventions.
NEVER:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→pass | 26,311 | 23,677 | -10% | 1 | 1 | 0% | 4,655 | 4,318 | -7% | 0 | 0 | — |
case-01 | fail→pass | 19,949 | 17,483 | -12% | 1 | 1 | 0% | 3,294 | 3,448 | +5% | 0 | 0 | — |
case-02 | fail→fail | 24,210 | 16,784 | -31% | 1 | 1 | 0% | 3,938 | 3,240 | -18% | 0 | 0 | — |
case-03 | fail→fail | 20,215 | 7,334 | -64% | 1 | 1 | 0% | 2,950 | 1,438 | -51% | 0 | 0 | — |
case-05 | fail→fail | 18,847 | 25,704 | +36% | 1 | 1 | 0% | 2,809 | 3,668 | +31% | 0 | 0 | — |
case-06 | fail→fail | 18,339 | 16,731 | -9% | 1 | 1 | 0% | 2,995 | 3,572 | +19% | 0 | 0 | — |
case-07 | fail→pass | 18,989 | 14,074 | -26% | 1 | 1 | 0% | 3,255 | 2,721 | -16% | 0 | 0 | — |
case-08 | fail→fail | 19,154 | 19,437 | +1% | 1 | 1 | 0% | 3,291 | 2,002 | -39% | 0 | 0 | — |
case-09 | fail→pass | 30,958 | 14,228 | -54% | 1 | 1 | 0% | 3,005 | 2,871 | -4% | 0 | 0 | — |
case-10 | fail→pass | 21,159 | 21,532 | +2% | 1 | 1 | 0% | 3,764 | 4,230 | +12% | 0 | 0 | — |
case-11 | fail→pass | 19,267 | 16,077 | -17% | 1 | 1 | 0% | 3,134 | 3,272 | +4% | 0 | 0 | — |
case-12 | fail→pass | 17,060 | 18,299 | +7% | 1 | 1 | 0% | 2,702 | 3,457 | +28% | 0 | 0 | — |
case-13 | fail→fail | 17,785 | 14,582 | -18% | 1 | 1 | 0% | 2,803 | 2,819 | +1% | 0 | 0 | — |
case-14 | fail→fail | 17,584 | 15,775 | -10% | 1 | 1 | 0% | 2,753 | 3,274 | +19% | 0 | 0 | — |
case-15 | fail→fail | 16,409 | 19,923 | +21% | 1 | 1 | 0% | 2,784 | 4,289 | +54% | 0 | 0 | — |
case-16 | fail→pass | 21,347 | 18,736 | -12% | 1 | 1 | 0% | 2,864 | 3,904 | +36% | 0 | 0 | — |
case-17 | fail→pass | 15,801 | 20,074 | +27% | 1 | 1 | 0% | 2,574 | 3,705 | +44% | 0 | 0 | — |
case-18 | fail→pass | 17,981 | 18,895 | +5% | 1 | 1 | 0% | 3,308 | 3,928 | +19% | 0 | 0 | — |
case-19 | fail→fail | 20,602 | 22,951 | +11% | 1 | 1 | 0% | 3,393 | 4,176 | +23% | 0 | 0 | — |
case-20 | pass→fail | 9,008 | 12,585 | +40% | 1 | 1 | 0% | 1,621 | 2,690 | +66% | 0 | 0 | — |
case-21 | pass→pass | 3,088 | 21,326 | +591% | 1 | 1 | 0% | 430 | 3,712 | +763% | 0 | 0 | — |
case-22 | pass→fail | 9,023 | 11,801 | +31% | 1 | 1 | 0% | 1,731 | 2,168 | +25% | 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 +36 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
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.