Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use this skill when first entering a new project or when asked to "get a basic understanding" of the codebase. It helps to map out the structure and locate documentation.
.claude/skills/miguelaperez-onboarding/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-20 | ✓→✓ | ＝ Same ✓ | 6% | 0% |
When you are tasked with understanding a new project, follow these steps to orient yourself:
README.md file. Read it to understand the project's purpose, setup, and core functionality.docs/ or documentation/ folder. Explore its contents to find architecture diagrams, API specs, or guides.list_dir on the project root to see high-level folders (e.g., src/, lib/, tests/, config/).package.json, requirements.txt, composer.json, go.mod).README.md exists, recommend creating one. You can use the templates located in resources/templates/:BASIC_README.md: For small projects or prototypes.STANDARD_README.md: For comprehensive project documentation.API_README.md: For documenting API-centric services.PULL_REQUEST_TEMPLATE.md: For standardizing pull request descriptions.docs/ folder, proactively suggest setting one up to keep the project organized.README.md or files in docs/) to reflect the changes.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | fail→fail | 11,080 | 6,624 | -40% | 1 | 1 | 0% | 1,972 | 1,565 | -21% | 0 | 0 | — |
case-20 | pass→pass | 14,065 | 12,754 | -9% | 1 | 1 | 0% | 3,094 | 3,291 | +6% | 0 | 0 | — |
case-01 | fail→fail | 2,535 | 1,686 | -33% | 1 | 1 | 0% | 464 | 710 | +53% | 0 | 0 | — |
case-02 | fail→fail | 7,029 | 1,708 | -76% | 1 | 1 | 0% | 1,279 | 668 | -48% | 0 | 0 | — |
case-03 | fail→fail | 2,642 | 1,703 | -36% | 1 | 1 | 0% | 331 | 670 | +102% | 0 | 0 | — |
case-04 | fail→pass | 11,313 | 5,981 | -47% | 1 | 1 | 0% | 1,936 | 1,493 | -23% | 0 | 0 | — |
case-05 | fail→fail | 9,421 | 6,018 | -36% | 1 | 1 | 0% | 1,722 | 1,554 | -10% | 0 | 0 | — |
case-06 | fail→pass | 12,527 | 4,270 | -66% | 1 | 1 | 0% | 2,486 | 1,217 | -51% | 0 | 0 | — |
case-08 | pass→pass | 8,212 | 4,082 | -50% | 1 | 1 | 0% | 1,537 | 1,110 | -28% | 0 | 0 | — |
case-09 | fail→pass | 9,527 | 4,122 | -57% | 1 | 1 | 0% | 1,603 | 1,036 | -35% | 0 | 0 | — |
case-10 | pass→pass | 9,707 | 5,265 | -46% | 1 | 1 | 0% | 1,907 | 1,402 | -26% | 0 | 0 | — |
case-11 | pass→pass | 5,949 | 3,102 | -48% | 1 | 1 | 0% | 1,057 | 946 | -11% | 0 | 0 | — |
case-12 | pass→pass | 11,955 | 10,776 | -10% | 1 | 1 | 0% | 2,171 | 2,143 | -1% | 0 | 0 | — |
case-13 | pass→pass | 4,375 | 1,903 | -57% | 1 | 1 | 0% | 767 | 694 | -10% | 0 | 0 | — |
case-14 | pass→pass | 4,780 | 1,807 | -62% | 1 | 1 | 0% | 858 | 640 | -25% | 0 | 0 | — |
case-15 | pass→pass | 8,510 | 2,824 | -67% | 1 | 1 | 0% | 1,431 | 910 | -36% | 0 | 0 | — |
case-16 | pass→pass | 8,924 | 2,501 | -72% | 1 | 1 | 0% | 1,659 | 822 | -50% | 0 | 0 | — |
case-17 | fail→pass | 7,683 | 1,436 | -81% | 1 | 1 | 0% | 1,343 | 627 | -53% | 0 | 0 | — |
case-18 | pass→pass | 9,209 | 4,340 | -53% | 1 | 1 | 0% | 1,529 | 1,022 | -33% | 0 | 0 | — |
case-19 | pass→pass | 7,175 | 1,864 | -74% | 1 | 1 | 0% | 1,215 | 680 | -44% | 0 | 0 | — |
case-21 | pass→pass | 6,786 | 6,196 | -9% | 1 | 1 | 0% | 1,277 | 1,539 | +21% | 0 | 0 | — |
case-22 | pass→pass | 13,830 | 11,824 | -15% | 1 | 1 | 0% | 2,473 | 2,370 | -4% | 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 +18 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.