Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Guides a new developer through five staged challenge sets covering architecture, domain, patterns, and hardening. Use when onboarding contributors.
.claude/skills/athola-onboard/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 190% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 182% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -58% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -4% | 0% |
Walk a new developer through the codebase in structured stages.
gauntlet:challenge)gauntlet:extract)| Stage | Focus | Categories | Difficulty | |-------|-------|------------|------------| | 1 | Big picture | architecture, data_flow | 1-2 | | 2 | Core domain | business_logic | 2-3 | | 3 | Interfaces | api_contract, data_flow | 3 | | 4 | Patterns | pattern, dependency | 3-4 | | 5 | Hardening | error_handling, business_logic | 4-5 |
After stage 5, the developer enters the regular gauntlet. Answer history carries over.
and challenge count survive session restarts and are loaded at Step 1
challenges in the current stage; partial completion does not advance
challenge before it is marked mastered
carries over into the regular gauntlet challenge pool without loss
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 7,411 | 11,349 | +53% | 1 | 1 | 0% | 1,215 | 1,992 | +64% | 0 | 0 | — |
case-02 | fail→pass | 4,480 | 10,447 | +133% | 1 | 1 | 0% | 698 | 2,026 | +190% | 0 | 0 | — |
case-03 | fail→pass | 4,579 | 10,205 | +123% | 1 | 1 | 0% | 720 | 2,032 | +182% | 0 | 0 | — |
case-04 | pass→pass | 6,188 | 3,681 | -41% | 1 | 1 | 0% | 980 | 792 | -19% | 0 | 0 | — |
case-05 | fail→pass | 16,166 | 4,365 | -73% | 1 | 1 | 0% | 2,583 | 1,095 | -58% | 0 | 0 | — |
case-06 | fail→pass | 5,680 | 3,131 | -45% | 1 | 1 | 0% | 881 | 844 | -4% | 0 | 0 | — |
case-07 | pass→pass | 12,779 | 3,046 | -76% | 1 | 1 | 0% | 2,047 | 854 | -58% | 0 | 0 | — |
case-20 | fail→pass | 3,064 | 1,481 | -52% | 1 | 1 | 0% | 446 | 637 | +43% | 0 | 0 | — |
case-08 | pass→pass | 7,273 | 2,454 | -66% | 1 | 1 | 0% | 1,131 | 850 | -25% | 0 | 0 | — |
case-09 | fail→pass | 7,323 | 1,579 | -78% | 1 | 1 | 0% | 1,047 | 607 | -42% | 0 | 0 | — |
case-10 | fail→pass | 7,528 | 1,908 | -75% | 1 | 1 | 0% | 1,203 | 701 | -42% | 0 | 0 | — |
case-11 | fail→pass | 9,837 | 2,074 | -79% | 1 | 1 | 0% | 1,570 | 744 | -53% | 0 | 0 | — |
case-12 | fail→pass | 6,548 | 3,606 | -45% | 1 | 1 | 0% | 1,061 | 970 | -9% | 0 | 0 | — |
case-13 | fail→pass | 6,347 | 2,963 | -53% | 1 | 1 | 0% | 942 | 802 | -15% | 0 | 0 | — |
case-14 | fail→pass | 8,056 | 2,414 | -70% | 1 | 1 | 0% | 1,186 | 716 | -40% | 0 | 0 | — |
case-15 | fail→pass | 6,722 | 2,179 | -68% | 1 | 1 | 0% | 965 | 719 | -25% | 0 | 0 | — |
case-16 | fail→pass | 13,367 | 1,649 | -88% | 1 | 1 | 0% | 2,071 | 607 | -71% | 0 | 0 | — |
case-17 | fail→pass | 11,348 | 1,812 | -84% | 1 | 1 | 0% | 1,680 | 672 | -60% | 0 | 0 | — |
case-18 | fail→pass | 9,341 | 4,168 | -55% | 1 | 1 | 0% | 1,386 | 1,071 | -23% | 0 | 0 | — |
case-19 | fail→pass | 9,068 | 2,714 | -70% | 1 | 1 | 0% | 1,352 | 839 | -38% | 0 | 0 | — |
case-21 | fail→pass | 8,458 | 2,006 | -76% | 1 | 1 | 0% | 1,268 | 694 | -45% | 0 | 0 | — |
case-22 | fail→pass | 10,205 | 2,227 | -78% | 1 | 1 | 0% | 1,549 | 753 | -51% | 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 +86 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.