Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Meta-skill — extract reusable skills from the current conversation. Identifies patterns that appeared 3+ times and proposes them as new skills. Pairs with skill-creator builtin.
.claude/skills/evolution-foundation-dev-learner/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -30% | 0% |
Derived from oh-my-claudecode (MIT, Yeachan Heo). Adapted for the EvoNexus Engineering Layer.
Meta-skill: scan the current conversation for patterns that appeared 3+ times and propose extracting them as reusable skills. The "self-improving workspace" feedback loop.
Scan the conversation for:
For each detected pattern:
Propose the new skill:
dev-{action} if engineering, {prefix}-{action} otherwise)If user accepts, hand off to skill-creator (builtin) to actually generate the skill file.
markdown## Learner Report — {session topic} ### Patterns Detected 1. **{Pattern name}** - Occurrences: 4 times - Steps: ... - Suggested skill: `dev-{action}` 2. **{Pattern name}** - ... ### Skill Proposals [Detailed proposal for each detected pattern] ### Recommendation [Which to extract first, which to skip]
skill-creator (builtin) — to generate the actual skill filecreate-agent (builtin) — if the pattern warrants a new agent instead| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→pass | 15,076 | 9,068 | -40% | 1 | 1 | 0% | 2,398 | 2,201 | -8% | 0 | 0 | — |
case-01 | fail→pass | 14,739 | 12,446 | -16% | 1 | 1 | 0% | 2,265 | 2,657 | +17% | 0 | 0 | — |
case-02 | fail→pass | 12,543 | 12,610 | +1% | 1 | 1 | 0% | 2,015 | 2,722 | +35% | 0 | 0 | — |
case-04 | pass→pass | 7,424 | 7,216 | -3% | 1 | 1 | 0% | 1,271 | 1,781 | +40% | 0 | 0 | — |
case-05 | fail→pass | 10,318 | 8,194 | -21% | 1 | 1 | 0% | 1,555 | 1,876 | +21% | 0 | 0 | — |
case-06 | fail→pass | 15,748 | 7,465 | -53% | 1 | 1 | 0% | 2,616 | 1,839 | -30% | 0 | 0 | — |
case-07 | fail→fail | 11,457 | 8,155 | -29% | 1 | 1 | 0% | 1,794 | 1,910 | +6% | 0 | 0 | — |
case-08 | pass→pass | 9,009 | 7,231 | -20% | 1 | 1 | 0% | 1,566 | 1,791 | +14% | 0 | 0 | — |
case-09 | fail→pass | 12,480 | 8,500 | -32% | 1 | 1 | 0% | 1,930 | 1,975 | +2% | 0 | 0 | — |
case-10 | pass→pass | 15,323 | 7,559 | -51% | 1 | 1 | 0% | 2,322 | 1,854 | -20% | 0 | 0 | — |
case-11 | fail→pass | 16,355 | 9,347 | -43% | 1 | 1 | 0% | 2,529 | 2,105 | -17% | 0 | 0 | — |
case-12 | pass→pass | 13,093 | 8,514 | -35% | 1 | 1 | 0% | 2,242 | 1,952 | -13% | 0 | 0 | — |
case-13 | pass→pass | 15,227 | 10,344 | -32% | 1 | 1 | 0% | 2,510 | 2,274 | -9% | 0 | 0 | — |
case-14 | pass→pass | 15,339 | 7,571 | -51% | 1 | 1 | 0% | 2,368 | 1,902 | -20% | 0 | 0 | — |
case-15 | pass→pass | 10,282 | 9,066 | -12% | 1 | 1 | 0% | 1,555 | 1,934 | +24% | 0 | 0 | — |
case-16 | pass→pass | 12,784 | 7,778 | -39% | 1 | 1 | 0% | 1,873 | 1,904 | +2% | 0 | 0 | — |
case-17 | fail→pass | 10,228 | 5,225 | -49% | 1 | 1 | 0% | 1,581 | 1,423 | -10% | 0 | 0 | — |
case-18 | fail→fail | 13,606 | 9,304 | -32% | 1 | 1 | 0% | 2,163 | 2,143 | -1% | 0 | 0 | — |
case-19 | fail→pass | 11,370 | 5,655 | -50% | 1 | 1 | 0% | 2,075 | 1,468 | -29% | 0 | 0 | — |
case-20 | fail→pass | 12,190 | 6,732 | -45% | 1 | 1 | 0% | 2,115 | 1,653 | -22% | 0 | 0 | — |
case-21 | fail→fail | 9,007 | 7,671 | -15% | 1 | 1 | 0% | 1,517 | 1,869 | +23% | 0 | 0 | — |
case-22 | fail→pass | 12,240 | 8,061 | -34% | 1 | 1 | 0% | 1,838 | 1,875 | +2% | 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 +50 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.