Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build a targeted stretching or mobility routine for the tightness you actually have — desk-stiff hips, a tight back, post-run legs — not a generic list. Use when asked for a stretching routine, my [back/hips/neck] is tight, mobility routine, or stretches for [activity]. Produces a short routine for the target area with hold times and cues, a warm-up vs recovery distinction, a daily-minimum version, and a plain 'ease in, don't force pain, see a pro for sharp/ongoing pain' note.
.claude/skills/mohitagw15856-stretching-routine/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 54% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -31% | 0% |
Generic "touch your toes" advice doesn't fix a specific tightness. This targets the area that's actually bothering you, gives the right kind of stretch for the moment (dynamic to warm up, static to recover), with clear hold times and cues — plus a two-minute version for the days you won't do the full thing.
Ask for these if not provided:
Type: dynamic warm-up / static recovery].
Daily 2-minute version: top 2–3 moves].
> Ease into each stretch — mild tension, not pain. Don't bounce. Sharp, sudden, or ongoing pain means stop and see a professional.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 23,636 | 16,759 | -29% | 1 | 1 | 0% | 2,557 | 3,062 | +20% | 0 | 0 | — |
case-02 | fail→fail | 20,624 | 19,874 | -4% | 1 | 1 | 0% | 2,679 | 3,161 | +18% | 0 | 0 | — |
case-03 | fail→fail | 21,927 | 11,988 | -45% | 1 | 1 | 0% | 2,890 | 2,927 | +1% | 0 | 0 | — |
case-04 | fail→pass | 18,967 | 16,101 | -15% | 1 | 1 | 0% | 2,159 | 2,564 | +19% | 0 | 0 | — |
case-05 | pass→fail | 20,591 | 19,909 | -3% | 1 | 1 | 0% | 2,415 | 2,713 | +12% | 0 | 0 | — |
case-06 | fail→fail | 15,819 | 18,939 | +20% | 1 | 1 | 0% | 2,396 | 3,057 | +28% | 0 | 0 | — |
case-07 | pass→pass | 15,774 | 17,758 | +13% | 1 | 1 | 0% | 1,651 | 2,513 | +52% | 0 | 0 | — |
case-08 | fail→pass | 18,163 | 14,946 | -18% | 1 | 1 | 0% | 2,105 | 2,531 | +20% | 0 | 0 | — |
case-09 | fail→fail | 19,327 | 19,140 | -1% | 1 | 1 | 0% | 2,282 | 2,702 | +18% | 0 | 0 | — |
case-10 | fail→fail | 17,932 | 18,015 | +0% | 1 | 1 | 0% | 2,139 | 2,527 | +18% | 0 | 0 | — |
case-11 | fail→fail | 18,886 | 22,012 | +17% | 1 | 1 | 0% | 2,380 | 3,143 | +32% | 0 | 0 | — |
case-12 | fail→fail | 19,369 | 17,010 | -12% | 1 | 1 | 0% | 2,235 | 2,720 | +22% | 0 | 0 | — |
case-13 | fail→fail | 17,797 | 14,844 | -17% | 1 | 1 | 0% | 1,923 | 2,567 | +33% | 0 | 0 | — |
case-14 | fail→pass | 15,023 | 16,273 | +8% | 1 | 1 | 0% | 1,687 | 2,605 | +54% | 0 | 0 | — |
case-15 | pass→pass | 17,000 | 16,587 | -2% | 1 | 1 | 0% | 1,961 | 2,793 | +42% | 0 | 0 | — |
case-16 | fail→fail | 21,708 | 13,275 | -39% | 1 | 1 | 0% | 1,824 | 2,588 | +42% | 0 | 0 | — |
case-17 | fail→pass | 15,217 | 20,545 | +35% | 1 | 1 | 0% | 2,372 | 2,766 | +17% | 0 | 0 | — |
case-18 | fail→fail | 17,079 | 12,870 | -25% | 1 | 1 | 0% | 1,923 | 2,639 | +37% | 0 | 0 | — |
case-19 | pass→pass | 11,051 | 16,659 | +51% | 1 | 1 | 0% | 1,686 | 2,608 | +55% | 0 | 0 | — |
case-20 | fail→pass | 20,163 | 8,492 | -58% | 1 | 1 | 0% | 2,969 | 2,047 | -31% | 0 | 0 | — |
case-21 | fail→fail | 20,129 | 26,593 | +32% | 1 | 1 | 0% | 3,699 | 5,194 | +40% | 0 | 0 | — |
case-22 | pass→pass | 14,012 | 9,788 | -30% | 1 | 1 | 0% | 1,994 | 2,478 | +24% | 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. 1 case got worse with the skill loaded, and it is 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.