Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user types /hep-graph or asks to create, list, inspect, or request a run of an Agentlas automation graph.
.claude/skills/agentlas-ai-hephaestus-graph/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 98% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -40% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -32% | 0% |
Saved graphs live in the local Agentlas database shared with Agentlas Desktop. This skill may create, list, inspect, or request a graph run; it does not execute the graph itself.
Use the first executable path:
bashCLI="" for candidate in \ "$(command -v agentlas 2>/dev/null)" \ "$HOME/.agentlas/runtime/current/bin/agentlas" \ "./bin/agentlas" do if [ -n "$candidate" ] && [ -x "$candidate" ]; then CLI="$candidate"; break; fi done [ -n "$CLI" ] || { echo "Agentlas CLI not found. Install it with: npm i -g agentlas" >&2; exit 1; }
Do not substitute Gemini CLI, Hephaestus routing, or Hub search when the Agentlas CLI is missing.
list: run "$CLI" graph list and report each graph'strigger, step count, and on/off state.
show <name>: run "$CLI" graph show "<name>". Preserve the tree wiring,external-effect and ask-first marks, branch sides, repeat edges, verification checklist, code steps, and any required input in the summary.
new <request>: run "$CLI" graph new "<request>" as an interview. Relayevery unanswered question and never invent schedule, external-effect, repetition, or checklist answers. The final save confirmation is the graph's approval. A new graph starts switched off.
run <name>: the user's direct run request is sufficient authority. Do notask for a second approval. Show the graph first; if it requires an input, obtain that missing value without inventing it. Then run "$CLI" graph run "<name>" -y, adding --input "<value>" when required.
Report exactly that a run was requested and will be picked up by Agentlas Desktop. Never claim the graph executed. On any non-zero exit, relay the exact refusal and stop; do not retry with a guess.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,438 | 9,347 | +26% | 1 | 1 | 0% | 808 | 972 | +20% | 0 | 0 | — |
case-02 | fail→fail | 7,579 | 13,100 | +73% | 1 | 1 | 0% | 1,141 | 1,025 | -10% | 0 | 0 | — |
case-03 | fail→fail | 5,542 | 15,995 | +189% | 1 | 1 | 0% | 741 | 1,039 | +40% | 0 | 0 | — |
case-04 | fail→pass | 17,352 | 5,433 | -69% | 1 | 1 | 0% | 633 | 1,255 | +98% | 0 | 0 | — |
case-05 | fail→pass | 14,015 | 3,501 | -75% | 1 | 1 | 0% | 2,180 | 1,010 | -54% | 0 | 0 | — |
case-06 | fail→fail | 11,479 | 13,854 | +21% | 1 | 1 | 0% | 1,432 | 972 | -32% | 0 | 0 | — |
case-07 | fail→fail | 4,063 | 13,116 | +223% | 1 | 1 | 0% | 523 | 1,080 | +107% | 0 | 0 | — |
case-08 | fail→pass | 10,070 | 2,707 | -73% | 1 | 1 | 0% | 1,348 | 803 | -40% | 0 | 0 | — |
case-09 | fail→fail | 11,845 | 5,351 | -55% | 1 | 1 | 0% | 1,677 | 1,024 | -39% | 0 | 0 | — |
case-10 | pass→pass | 7,809 | 3,633 | -53% | 1 | 1 | 0% | 1,131 | 941 | -17% | 0 | 0 | — |
case-11 | fail→fail | 8,080 | 4,778 | -41% | 1 | 1 | 0% | 1,095 | 894 | -18% | 0 | 0 | — |
case-12 | pass→fail | 3,970 | 12,066 | +204% | 1 | 1 | 0% | 474 | 1,342 | +183% | 0 | 0 | — |
case-13 | fail→pass | 12,281 | 5,887 | -52% | 1 | 1 | 0% | 1,908 | 1,299 | -32% | 0 | 0 | — |
case-14 | fail→pass | 9,431 | 2,963 | -69% | 1 | 1 | 0% | 1,276 | 865 | -32% | 0 | 0 | — |
case-15 | fail→pass | 11,247 | 4,087 | -64% | 1 | 1 | 0% | 1,537 | 977 | -36% | 0 | 0 | — |
case-16 | fail→fail | 12,920 | 9,243 | -28% | 1 | 1 | 0% | 1,837 | 940 | -49% | 0 | 0 | — |
case-17 | fail→fail | 16,775 | 9,781 | -42% | 1 | 1 | 0% | 2,339 | 828 | -65% | 0 | 0 | — |
case-18 | fail→pass | 4,282 | 10,529 | +146% | 1 | 1 | 0% | 494 | 2,146 | +334% | 0 | 0 | — |
case-19 | fail→pass | 8,409 | 10,090 | +20% | 1 | 1 | 0% | 1,157 | 1,944 | +68% | 0 | 0 | — |
case-20 | fail→fail | 16,423 | 12,118 | -26% | 1 | 1 | 0% | 2,617 | 958 | -63% | 0 | 0 | — |
case-21 | pass→pass | 11,001 | 3,399 | -69% | 1 | 1 | 0% | 1,720 | 1,075 | -38% | 0 | 0 | — |
case-22 | fail→fail | 14,406 | 11,234 | -22% | 1 | 1 | 0% | 1,616 | 1,087 | -33% | 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 12 counted toward the lift figure. The other 10 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 +32 percentage points is the difference between those two pass rates over the 12 comparable cases. 2 cases got worse with the skill loaded, and they are 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.