Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Search Agentlas Cloud and Hub candidates without invoking agents.
.claude/skills/agentlas-ai-hep-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -51% | 0% |
Update fallback: 자동 업데이트가 안 되면 hephaestus update를 한 번 실행하세요. 업데이트하지 않아도 현재 버전 명령은 그대로 동작합니다.
Raw arguments: everything the user typed after /skill:hep-search.
Codex plugins cannot register slash commands, so this custom prompt is the explicit entrypoint: /prompts:hep-search.
bashRUNNER="" for c in "$HOME/.agentlas/runtime/current/bin/hephaestus" ./bin/hephaestus; do [ -x "$c" ] && RUNNER="$c" && break done [ -n "$RUNNER" ] || { echo "Hephaestus runtime not found. Run the installer first." >&2; exit 1; } if [ "${HEPHAESTUS_AUTH_AUTOPOPUP:-1}" != "0" ]; then "$RUNNER" auth ensure --timeout 180 >/dev/null 2>&1 || true fi "$RUNNER" search "$ARGUMENTS" --runtime codex --limit 10
cloud: my own Agentlas Cloud packages.hub: public Agentlas Hub marketplace.Do not invoke agents from this prompt. Use /prompts:hep-call next when the user names exact slugs.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,816 | 10,836 | -15% | 1 | 1 | 0% | 1,281 | 710 | -45% | 0 | 0 | — |
case-02 | fail→fail | 21,047 | 17,521 | -17% | 1 | 1 | 0% | 2,590 | 904 | -65% | 0 | 0 | — |
case-03 | fail→fail | 9,889 | 11,683 | +18% | 1 | 1 | 0% | 1,435 | 801 | -44% | 0 | 0 | — |
case-04 | pass→pass | 8,779 | 5,243 | -40% | 1 | 1 | 0% | 1,103 | 1,016 | -8% | 0 | 0 | — |
case-05 | fail→fail | 61,549 | 5,478 | -91% | 1 | 1 | 0% | 5,920 | 1,475 | -75% | 0 | 0 | — |
case-06 | fail→fail | 15,025 | 9,018 | -40% | 1 | 1 | 0% | 2,234 | 910 | -59% | 0 | 0 | — |
case-07 | fail→pass | 8,563 | 13,701 | +60% | 1 | 1 | 0% | 1,261 | 1,093 | -13% | 0 | 0 | — |
case-08 | fail→fail | 22,348 | 3,910 | -83% | 1 | 1 | 0% | 1,035 | 871 | -16% | 0 | 0 | — |
case-09 | fail→fail | 4,554 | 17,579 | +286% | 1 | 1 | 0% | 679 | 1,493 | +120% | 0 | 0 | — |
case-10 | fail→pass | 14,141 | 6,705 | -53% | 1 | 1 | 0% | 1,752 | 1,322 | -25% | 0 | 0 | — |
case-11 | fail→pass | 9,730 | 3,025 | -69% | 1 | 1 | 0% | 1,624 | 776 | -52% | 0 | 0 | — |
case-12 | pass→pass | 8,711 | 3,346 | -62% | 1 | 1 | 0% | 1,356 | 829 | -39% | 0 | 0 | — |
case-13 | fail→fail | 11,156 | 3,096 | -72% | 1 | 1 | 0% | 1,527 | 667 | -56% | 0 | 0 | — |
case-14 | fail→pass | 13,900 | 6,442 | -54% | 1 | 1 | 0% | 2,088 | 1,396 | -33% | 0 | 0 | — |
case-15 | fail→pass | 23,775 | 5,123 | -78% | 1 | 1 | 0% | 2,223 | 1,095 | -51% | 0 | 0 | — |
case-16 | fail→pass | 9,447 | 5,115 | -46% | 1 | 1 | 0% | 1,272 | 1,065 | -16% | 0 | 0 | — |
case-17 | fail→pass | 14,954 | 6,438 | -57% | 1 | 1 | 0% | 2,001 | 1,286 | -36% | 0 | 0 | — |
case-18 | fail→pass | 18,579 | 1,931 | -90% | 1 | 1 | 0% | 2,761 | 517 | -81% | 0 | 0 | — |
case-19 | pass→pass | 13,781 | 6,660 | -52% | 1 | 1 | 0% | 1,764 | 1,431 | -19% | 0 | 0 | — |
case-20 | fail→pass | 7,950 | 5,227 | -34% | 1 | 1 | 0% | 827 | 1,227 | +48% | 0 | 0 | — |
case-21 | pass→pass | 7,656 | 3,111 | -59% | 1 | 1 | 0% | 1,068 | 729 | -32% | 0 | 0 | — |
case-22 | fail→fail | 9,869 | 10,109 | +2% | 1 | 1 | 0% | 1,753 | 1,930 | +10% | 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, and 17 counted toward the lift figure. The other 5 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 +41 percentage points is the difference between those two pass rates over the 17 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.