Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Fix the screen-hunch with a realistic plan — the desk fixes, the two or three exercises that counter it, and movement habits that beat any single stretch. Use when asked how to fix my posture, I have bad posture from sitting, tech neck, or rounded shoulders help. Produces a quick posture-cause read, immediate desk/setup fixes, a few high-value strengthening and mobility moves, movement-break habits, and honest expectations — plus a 'see a professional for pain/numbness' flag.
.claude/skills/mohitagw15856-posture-reset-plan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 50% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 18% | 0% |
"Sit up straight" doesn't work because posture isn't willpower — it's a setup and a habit. This tackles the real drivers of the desk-hunch: an ergonomics fix so your body isn't fighting the chair, a couple of exercises to strengthen what's weak and open what's tight, and movement breaks — because the best posture is the next position, not one perfect pose.
Ask for these if not provided:
Likely driver: forward head / rounded shoulders / …].
Fix your setup now
Do these (daily)
Movement habit: break every X min — a specific cue]. Expect: weeks of consistency. See a pro if: pain, numbness, or tingling.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 18,088 | 15,014 | -17% | 1 | 1 | 0% | 1,927 | 2,355 | +22% | 0 | 0 | — |
case-01 | fail→pass | 25,059 | 16,649 | -34% | 1 | 1 | 0% | 3,205 | 2,797 | -13% | 0 | 0 | — |
case-02 | pass→pass | 23,434 | 16,587 | -29% | 1 | 1 | 0% | 2,894 | 2,824 | -2% | 0 | 0 | — |
case-03 | pass→pass | 23,763 | 16,555 | -30% | 1 | 1 | 0% | 3,028 | 2,944 | -3% | 0 | 0 | — |
case-04 | pass→pass | 17,850 | 14,043 | -21% | 1 | 1 | 0% | 1,864 | 2,237 | +20% | 0 | 0 | — |
case-05 | pass→pass | 30,690 | 23,242 | -24% | 1 | 1 | 0% | 3,401 | 3,782 | +11% | 0 | 0 | — |
case-07 | fail→pass | 18,416 | 19,786 | +7% | 1 | 1 | 0% | 2,020 | 3,030 | +50% | 0 | 0 | — |
case-08 | fail→pass | 28,694 | 20,187 | -30% | 1 | 1 | 0% | 2,122 | 2,919 | +38% | 0 | 0 | — |
case-09 | pass→pass | 22,989 | 15,398 | -33% | 1 | 1 | 0% | 2,485 | 3,219 | +30% | 0 | 0 | — |
case-10 | fail→pass | 21,594 | 19,024 | -12% | 1 | 1 | 0% | 2,118 | 2,620 | +24% | 0 | 0 | — |
case-11 | pass→pass | 15,596 | 17,427 | +12% | 1 | 1 | 0% | 1,927 | 2,741 | +42% | 0 | 0 | — |
case-12 | pass→pass | 17,775 | 13,428 | -24% | 1 | 1 | 0% | 2,687 | 2,552 | -5% | 0 | 0 | — |
case-13 | pass→pass | 18,791 | 24,600 | +31% | 1 | 1 | 0% | 2,121 | 3,224 | +52% | 0 | 0 | — |
case-14 | fail→pass | 22,708 | 11,868 | -48% | 1 | 1 | 0% | 2,269 | 2,673 | +18% | 0 | 0 | — |
case-15 | fail→pass | 10,788 | 17,269 | +60% | 1 | 1 | 0% | 1,668 | 2,502 | +50% | 0 | 0 | — |
case-16 | pass→pass | 24,211 | 22,249 | -8% | 1 | 1 | 0% | 2,494 | 2,971 | +19% | 0 | 0 | — |
case-17 | fail→pass | 25,510 | 18,891 | -26% | 1 | 1 | 0% | 2,618 | 2,756 | +5% | 0 | 0 | — |
case-18 | pass→pass | 17,332 | 17,191 | -1% | 1 | 1 | 0% | 2,323 | 2,838 | +22% | 0 | 0 | — |
case-19 | pass→pass | 115,323 | 18,362 | -84% | 1 | 1 | 0% | 2,225 | 2,610 | +17% | 0 | 0 | — |
case-20 | pass→pass | 23,499 | 26,850 | +14% | 1 | 1 | 0% | 2,276 | 2,866 | +26% | 0 | 0 | — |
case-21 | fail→pass | 15,356 | 12,155 | -21% | 1 | 1 | 0% | 2,339 | 2,742 | +17% | 0 | 0 | — |
case-22 | fail→pass | 35,872 | 11,163 | -69% | 1 | 1 | 0% | 2,287 | 2,560 | +12% | 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 +41 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.