Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user types /agentlas-graph, /agentlas graph, or /hep-graph to create, list, inspect, or run Agentlas automation graphs.
.claude/skills/agentlas-ai-agentlas-graph/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | 107% | 0% |
| case-17 | ✓→✗ | ▼ Worse | -4% | 0% |
| case-20 | ✓→✓ | ＝ Same ✓ | 37% | 0% |
| case-21 | ✓→✓ | ＝ Same ✓ | 9% | 0% |
| case-22 | ✓→✓ | ＝ Same ✓ | 68% | 0% |
Identical to /skill:hep-graph and /agentlas graph <request>. Locate the sibling skill directory hep-graph (i.e. ../hep-graph/SKILL.md relative to this file, under the same kimi/skills/ root this skill was loaded from), read its SKILL.md, and follow its instructions exactly — treating everything typed after /skill:agentlas-graph as that command's request.
Do not improvise a separate workflow and do not summarize hep-graph/SKILL.md from memory; that file is the sole authority for this command's behavior.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 11,666 | 24,972 | +114% | 1 | 1 | 0% | 2,137 | 4,672 | +119% | 0 | 0 | — |
case-02 | fail→fail | 56,752 | 8,049 | -86% | 1 | 1 | 0% | 2,995 | 557 | -81% | 0 | 0 | — |
case-03 | fail→fail | 33,240 | 14,978 | -55% | 1 | 1 | 0% | 2,560 | 2,475 | -3% | 0 | 0 | — |
case-04 | fail→fail | 67,865 | 60,044 | -12% | 1 | 1 | 0% | 7,461 | 6,457 | -13% | 0 | 0 | — |
case-05 | fail→fail | 27,645 | 79,372 | +187% | 1 | 1 | 0% | 5,219 | 8,361 | +60% | 0 | 0 | — |
case-06 | fail→fail | 16,249 | 14,182 | -13% | 1 | 1 | 0% | 2,621 | 2,172 | -17% | 0 | 0 | — |
case-07 | fail→fail | 42,388 | 65,916 | +56% | 1 | 1 | 0% | 8,210 | 8,359 | +2% | 0 | 0 | — |
case-08 | fail→fail | 39,660 | 45,186 | +14% | 1 | 1 | 0% | 5,315 | 8,358 | +57% | 0 | 0 | — |
case-09 | fail→fail | 45,582 | 42,698 | -6% | 1 | 1 | 0% | 3,843 | 8,356 | +117% | 0 | 0 | — |
case-10 | fail→fail | 24,072 | 20,699 | -14% | 1 | 1 | 0% | 3,226 | 3,458 | +7% | 0 | 0 | — |
case-11 | fail→fail | 21,749 | 22,510 | +3% | 1 | 1 | 0% | 3,412 | 4,018 | +18% | 0 | 0 | — |
case-12 | fail→fail | 41,138 | 45,200 | +10% | 1 | 1 | 0% | 7,784 | 8,358 | +7% | 0 | 0 | — |
case-13 | fail→fail | 25,502 | 20,722 | -19% | 1 | 1 | 0% | 3,613 | 3,426 | -5% | 0 | 0 | — |
case-14 | fail→pass | 26,751 | 40,929 | +53% | 1 | 1 | 0% | 4,165 | 8,638 | +107% | 0 | 0 | — |
case-15 | fail→fail | 21,803 | 67,092 | +208% | 1 | 1 | 0% | 3,023 | 4,925 | +63% | 0 | 0 | — |
case-16 | fail→fail | 21,539 | 24,769 | +15% | 1 | 1 | 0% | 3,354 | 3,220 | -4% | 0 | 0 | — |
case-17 | pass→fail | 25,867 | 26,416 | +2% | 1 | 1 | 0% | 5,274 | 5,067 | -4% | 0 | 0 | — |
case-18 | fail→fail | 18,230 | 22,484 | +23% | 1 | 1 | 0% | 2,830 | 3,616 | +28% | 0 | 0 | — |
case-19 | fail→fail | 28,065 | 48,591 | +73% | 1 | 1 | 0% | 5,302 | 9,536 | +80% | 0 | 0 | — |
case-20 | pass→pass | 12,792 | 17,846 | +40% | 1 | 1 | 0% | 2,810 | 3,858 | +37% | 0 | 0 | — |
case-21 | pass→pass | 24,528 | 24,046 | -2% | 1 | 1 | 0% | 4,167 | 4,554 | +9% | 0 | 0 | — |
case-22 | pass→pass | 4,828 | 6,588 | +36% | 1 | 1 | 0% | 694 | 1,165 | +68% | 0 | 0 | — |
case-23 | pass→pass | 27,664 | 18,012 | -35% | 1 | 1 | 0% | 3,091 | 2,890 | -7% | 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. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of 0 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.