Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build durable backend AI agents with the eve framework. Use when creating, editing, or debugging an eve project — agent instructions, skills, tools, connections, channels, sandboxes, subagents, schedules, or evals.
.claude/skills/asymmetric-al-eve/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -63% | 0% |
eve is a filesystem-first framework for durable backend AI agents. An agent is a directory on disk — instructions, skills, tools, connections, channels, subagents, and schedules are all files — and eve compiles and runs it.
The complete documentation ships inside the eve package. Do not rely on this skill for guidance — always read the bundled docs, which match the installed version exactly:
node_modules/eve/docs/Start with node_modules/eve/docs/README.md. It contains the full index and recommended reading order. Before writing any eve code, read the relevant guide there first.
If eve is not installed yet, install it (npm install eve) or scaffold a new agent with npx eve init <agent-name>, then read the bundled docs.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | fail→pass | 14,477 | 5,667 | -61% | 1 | 1 | 0% | 2,731 | 1,291 | -53% | 0 | 0 | — |
case-01 | fail→pass | 10,738 | 6,189 | -42% | 1 | 1 | 0% | 1,759 | 1,222 | -31% | 0 | 0 | — |
case-02 | fail→pass | 10,892 | 3,732 | -66% | 1 | 1 | 0% | 1,784 | 805 | -55% | 0 | 0 | — |
case-03 | fail→pass | 14,439 | 3,909 | -73% | 1 | 1 | 0% | 2,117 | 763 | -64% | 0 | 0 | — |
case-04 | fail→pass | 8,782 | 2,267 | -74% | 1 | 1 | 0% | 1,376 | 515 | -63% | 0 | 0 | — |
case-05 | fail→pass | 9,201 | 2,177 | -76% | 1 | 1 | 0% | 1,499 | 526 | -65% | 0 | 0 | — |
case-06 | fail→pass | 17,583 | 8,588 | -51% | 1 | 1 | 0% | 2,842 | 1,508 | -47% | 0 | 0 | — |
case-07 | pass→pass | 13,941 | 4,769 | -66% | 1 | 1 | 0% | 2,086 | 901 | -57% | 0 | 0 | — |
case-13 | fail→pass | 5,465 | 2,378 | -56% | 1 | 1 | 0% | 892 | 662 | -26% | 0 | 0 | — |
case-08 | fail→pass | 15,333 | 3,109 | -80% | 1 | 1 | 0% | 2,342 | 660 | -72% | 0 | 0 | — |
case-09 | pass→pass | 18,332 | 4,165 | -77% | 1 | 1 | 0% | 2,811 | 845 | -70% | 0 | 0 | — |
case-10 | fail→pass | 5,385 | 1,851 | -66% | 1 | 1 | 0% | 1,082 | 527 | -51% | 0 | 0 | — |
case-11 | fail→pass | 8,375 | 1,469 | -82% | 1 | 1 | 0% | 1,479 | 464 | -69% | 0 | 0 | — |
case-14 | pass→pass | 13,343 | 8,638 | -35% | 1 | 1 | 0% | 2,421 | 1,797 | -26% | 0 | 0 | — |
case-15 | fail→pass | 6,378 | 1,290 | -80% | 1 | 1 | 0% | 1,080 | 390 | -64% | 0 | 0 | — |
case-16 | pass→pass | 12,370 | 3,317 | -73% | 1 | 1 | 0% | 2,460 | 859 | -65% | 0 | 0 | — |
case-17 | fail→pass | 8,281 | 1,972 | -76% | 1 | 1 | 0% | 1,667 | 560 | -66% | 0 | 0 | — |
case-18 | fail→pass | 8,881 | 2,368 | -73% | 1 | 1 | 0% | 1,376 | 556 | -60% | 0 | 0 | — |
case-19 | fail→pass | 14,979 | 2,656 | -82% | 1 | 1 | 0% | 2,658 | 699 | -74% | 0 | 0 | — |
case-20 | pass→pass | 14,468 | 10,294 | -29% | 1 | 1 | 0% | 2,881 | 2,319 | -20% | 0 | 0 | — |
case-21 | pass→pass | 11,732 | 9,927 | -15% | 1 | 1 | 0% | 2,344 | 2,141 | -9% | 0 | 0 | — |
case-22 | pass→pass | 10,430 | 5,802 | -44% | 1 | 1 | 0% | 1,873 | 1,278 | -32% | 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 +68 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.