Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →High-Performance CLI AI Agents (Claude Code, Gemini CLI, Open Code).
.claude/skills/majiayu000-terminal-agents/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -1% | 0% |
> !TIP] > These tools allow running AI Agents directly in the terminal, bypassing browser limitations and enabling direct file system access. > Source: "You've Been Using AI the Hard Way" (Transcript).
Role: Premium Autonomous Developer.
claude.md for persistent project context.claude (interactive) or claude -p "prompt" (headless).Role: Accessible & Fast Terminal AI.
gemini.md.Role: Private & Flexible.
config.jsonc.Strategy: Run them all in the same directory.
claude.md, gemini.md, and agents.md (Codex) are synced or referenced.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,706 | 18,262 | -16% | 1 | 1 | 0% | 3,523 | 2,739 | -22% | 0 | 0 | — |
case-02 | fail→fail | 19,228 | 29,139 | +52% | 1 | 1 | 0% | 3,574 | 4,631 | +30% | 0 | 0 | — |
case-03 | fail→fail | 25,991 | 23,620 | -9% | 1 | 1 | 0% | 4,473 | 3,570 | -20% | 0 | 0 | — |
case-04 | pass→pass | 6,039 | 3,050 | -49% | 1 | 1 | 0% | 1,008 | 987 | -2% | 0 | 0 | — |
case-05 | pass→pass | 20,911 | 19,302 | -8% | 1 | 1 | 0% | 2,590 | 2,804 | +8% | 0 | 0 | — |
case-06 | pass→pass | 4,198 | 2,623 | -38% | 1 | 1 | 0% | 670 | 850 | +27% | 0 | 0 | — |
case-07 | pass→pass | 13,431 | 12,105 | -10% | 1 | 1 | 0% | 1,736 | 1,459 | -16% | 0 | 0 | — |
case-08 | pass→pass | 12,534 | 2,033 | -84% | 1 | 1 | 0% | 1,215 | 851 | -30% | 0 | 0 | — |
case-09 | fail→pass | 29,582 | 10,425 | -65% | 1 | 1 | 0% | 2,686 | 1,391 | -48% | 0 | 0 | — |
case-10 | fail→pass | 16,804 | 11,578 | -31% | 1 | 1 | 0% | 1,725 | 1,647 | -5% | 0 | 0 | — |
case-11 | pass→pass | 20,992 | 15,389 | -27% | 1 | 1 | 0% | 2,379 | 2,234 | -6% | 0 | 0 | — |
case-12 | pass→pass | 16,842 | 13,759 | -18% | 1 | 1 | 0% | 2,597 | 2,811 | +8% | 0 | 0 | — |
case-13 | fail→pass | 9,770 | 7,749 | -21% | 1 | 1 | 0% | 1,530 | 950 | -38% | 0 | 0 | — |
case-14 | fail→fail | 44,526 | 40,103 | -10% | 1 | 1 | 0% | 2,037 | 2,656 | +30% | 0 | 0 | — |
case-15 | pass→pass | 17,550 | 7,564 | -57% | 1 | 1 | 0% | 2,033 | 975 | -52% | 0 | 0 | — |
case-16 | pass→pass | 17,544 | 5,873 | -67% | 1 | 1 | 0% | 1,806 | 1,148 | -36% | 0 | 0 | — |
case-17 | fail→pass | 17,246 | 17,742 | +3% | 1 | 1 | 0% | 2,613 | 2,599 | -1% | 0 | 0 | — |
case-18 | pass→pass | 9,801 | 2,439 | -75% | 1 | 1 | 0% | 1,638 | 914 | -44% | 0 | 0 | — |
case-19 | fail→pass | 15,796 | 6,061 | -62% | 1 | 1 | 0% | 2,101 | 1,451 | -31% | 0 | 0 | — |
case-20 | pass→pass | 11,132 | 17,235 | +55% | 1 | 1 | 0% | 2,016 | 2,610 | +29% | 0 | 0 | — |
case-21 | pass→pass | 11,389 | 19,477 | +71% | 1 | 1 | 0% | 2,155 | 2,871 | +33% | 0 | 0 | — |
case-22 | pass→pass | 13,878 | 14,196 | +2% | 1 | 1 | 0% | 2,352 | 2,993 | +27% | 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 +27 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.