Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Core skill activation protocol - establishes mandatory workflows for finding and applying skills before any task
.claude/skills/dicklesworthstone-using-superpowers/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -6% | 0% |
> "If a skill exists for your task, you MUST use it. No exceptions. No rationalizations."
Skills encode proven solutions. Ignoring them means repeating solved problems.
Before ANY action, complete this mental checklist:
□ What skills might apply here?
□ Have I checked the skill registry?
□ Did I read the CURRENT version (not memory)?
□ Am I following it exactly?If a relevant skill exists → USE IT.
"I'm using [Skill Name] to [what you're doing]."If a skill has a checklist → Create todos for EACH item. Don't work through checklists mentally. Track them.
User says "Fix X" → That's the WHAT, not the HOW. Still follow brainstorming, TDD, verification workflows.
| Category | When to Use | |----------|-------------| | Checkers | Before assumptions, before actions | | Collaboration | Planning, brainstorming, executing plans | | Debugging | When something's broken | | Problem-Solving | When stuck | | Testing | TDD, test design | | Meta | Writing/improving skills |
Starting a feature → brainstorming → writing-plans → executing-plans
Something broken → systematic-debugging or root-cause-tracing
About to assume → assumption-checker
About to modify → pre-action-verifier
Claiming done → verification-before-completion| Thought | Reality | |---------|---------| | "I remember this skill" | Skills evolve. Read current version. | | "This doesn't need a skill" | If skill exists, use it. | | "Let me just do this first" | Check skills BEFORE anything. | | "This is simple" | Simple tasks become complex. Use skills. |
Before taking action, the checker agents validate:
assumption-checker → Are you guessing?context-validator → Do you have enough context?intent-clarifier → Is the request clear?pre-action-verifier → Are prerequisites met?fact-checker → Are your claims about code accurate?These checkers are strict mode - they halt and ask rather than proceeding.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 14,245 | 22,053 | +55% | 1 | 1 | 0% | 2,373 | 3,408 | +44% | 0 | 0 | — |
case-02 | fail→fail | 4,009 | 13,893 | +247% | 1 | 1 | 0% | 529 | 2,686 | +408% | 0 | 0 | — |
case-03 | pass→pass | 10,087 | 15,020 | +49% | 1 | 1 | 0% | 1,481 | 3,161 | +113% | 0 | 0 | — |
case-04 | pass→pass | 4,395 | 12,684 | +189% | 1 | 1 | 0% | 657 | 2,532 | +285% | 0 | 0 | — |
case-05 | pass→pass | 3,280 | 4,683 | +43% | 1 | 1 | 0% | 624 | 1,494 | +139% | 0 | 0 | — |
case-06 | pass→pass | 6,930 | 5,284 | -24% | 1 | 1 | 0% | 1,050 | 1,464 | +39% | 0 | 0 | — |
case-07 | fail→pass | 12,335 | 5,150 | -58% | 1 | 1 | 0% | 1,685 | 1,509 | -10% | 0 | 0 | — |
case-08 | pass→pass | 11,496 | 4,033 | -65% | 1 | 1 | 0% | 1,924 | 1,161 | -40% | 0 | 0 | — |
case-09 | pass→pass | 13,772 | 3,096 | -78% | 1 | 1 | 0% | 1,771 | 1,199 | -32% | 0 | 0 | — |
case-10 | fail→pass | 13,816 | 6,048 | -56% | 1 | 1 | 0% | 2,158 | 1,494 | -31% | 0 | 0 | — |
case-11 | fail→fail | 10,406 | 4,420 | -58% | 1 | 1 | 0% | 1,596 | 1,350 | -15% | 0 | 0 | — |
case-12 | fail→pass | 10,587 | 1,981 | -81% | 1 | 1 | 0% | 1,681 | 969 | -42% | 0 | 0 | — |
case-13 | fail→pass | 5,700 | 2,518 | -56% | 1 | 1 | 0% | 879 | 932 | +6% | 0 | 0 | — |
case-14 | pass→pass | 7,999 | 2,838 | -65% | 1 | 1 | 0% | 1,145 | 1,061 | -7% | 0 | 0 | — |
case-15 | pass→fail | 11,058 | 3,209 | -71% | 1 | 1 | 0% | 1,688 | 1,112 | -34% | 0 | 0 | — |
case-16 | pass→pass | 14,119 | 8,573 | -39% | 1 | 1 | 0% | 1,795 | 1,930 | +8% | 0 | 0 | — |
case-17 | pass→pass | 4,861 | 2,443 | -50% | 1 | 1 | 0% | 728 | 1,087 | +49% | 0 | 0 | — |
case-18 | pass→pass | 11,734 | 4,314 | -63% | 1 | 1 | 0% | 1,812 | 1,383 | -24% | 0 | 0 | — |
case-19 | pass→pass | 7,454 | 3,586 | -52% | 1 | 1 | 0% | 1,099 | 1,202 | +9% | 0 | 0 | — |
case-20 | fail→pass | 13,284 | 6,614 | -50% | 1 | 1 | 0% | 1,703 | 1,601 | -6% | 0 | 0 | — |
case-21 | fail→pass | 8,094 | 6,532 | -19% | 1 | 1 | 0% | 1,080 | 1,796 | +66% | 0 | 0 | — |
case-22 | pass→pass | 12,502 | 2,751 | -78% | 1 | 1 | 0% | 1,655 | 1,159 | -30% | 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 +23 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.