Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Configure Cobbler model routes and external-agent preferences for this checkout. Use when the user types /setup-cobbler, asks to set up Cobbler model routes, onboard models, or update which tools handle planning/implement/review in Claude Code.
.claude/skills/aigorahub-setup-cobbler/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -26% | 0% |
<!-- elves-managed-alias: claude-skill-alias v1 -->
This is an Elves-managed Claude Code alias for /setup-cobbler.
Run the model onboarding contract from the main elves skill and references/model-onboarding.md. Operator CLI:
bashpython3 scripts/cobbler_agents.py onboard plan|show|apply|probe [--json] python3 scripts/cobbler_agents.py setup … # apply-only / inventory
Setup must not require OpenRouter or any external provider key. Native-only Elves remains fully usable without running setup.
onboard plan --json — inventory tools, detect env names present (never values),emit purpose→route questions.
optional math evolutionary search). Offer host-native first. Respect available CLIs/keys.
onboard apply writes ignored .elves/models.toml only. Never stage it. Never pasteAPI keys into TOML, chat, or the Survival Guide.
onboard probe runs structural checks (PATH, --help, env names). Optional--smoke only if the user wants a paid live check — host runs a real tiny completion; fake smoke does not count.
show → re-interview → apply --force → probe).Codex equivalent (not a top-level slash command): $elves setup-cobbler or natural language such as "Set up Cobbler external-agent preferences" / "onboard my models."
Compatibility: /setup-council is the same setup contract with a Council-era name.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,847 | 15,126 | +9% | 1 | 1 | 0% | 2,676 | 683 | -74% | 0 | 0 | — |
case-02 | fail→fail | 22,733 | 2,961 | -87% | 1 | 1 | 0% | 2,677 | 793 | -70% | 0 | 0 | — |
case-03 | fail→fail | 11,736 | 10,281 | -12% | 1 | 1 | 0% | 1,765 | 686 | -61% | 0 | 0 | — |
case-04 | pass→pass | 13,373 | 9,156 | -32% | 1 | 1 | 0% | 2,546 | 2,034 | -20% | 0 | 0 | — |
case-05 | pass→pass | 9,095 | 7,596 | -16% | 1 | 1 | 0% | 1,693 | 1,947 | +15% | 0 | 0 | — |
case-06 | pass→pass | 10,407 | 10,202 | -2% | 1 | 1 | 0% | 1,869 | 2,265 | +21% | 0 | 0 | — |
case-07 | fail→pass | 8,934 | 3,062 | -66% | 1 | 1 | 0% | 1,525 | 1,020 | -33% | 0 | 0 | — |
case-08 | fail→pass | 13,978 | 4,825 | -65% | 1 | 1 | 0% | 1,850 | 1,280 | -31% | 0 | 0 | — |
case-09 | fail→pass | 17,721 | 6,696 | -62% | 1 | 1 | 0% | 2,882 | 1,496 | -48% | 0 | 0 | — |
case-10 | fail→pass | 10,918 | 5,402 | -51% | 1 | 1 | 0% | 1,880 | 1,351 | -28% | 0 | 0 | — |
case-11 | fail→pass | 7,871 | 3,506 | -55% | 1 | 1 | 0% | 1,393 | 1,037 | -26% | 0 | 0 | — |
case-12 | fail→pass | 14,846 | 4,333 | -71% | 1 | 1 | 0% | 2,099 | 1,010 | -52% | 0 | 0 | — |
case-13 | fail→pass | 8,445 | 5,274 | -38% | 1 | 1 | 0% | 1,444 | 1,380 | -4% | 0 | 0 | — |
case-14 | fail→pass | 13,581 | 9,189 | -32% | 1 | 1 | 0% | 1,731 | 1,251 | -28% | 0 | 0 | — |
case-15 | fail→pass | 11,614 | 5,600 | -52% | 1 | 1 | 0% | 1,757 | 1,482 | -16% | 0 | 0 | — |
case-16 | pass→pass | 10,616 | 5,452 | -49% | 1 | 1 | 0% | 1,815 | 1,315 | -28% | 0 | 0 | — |
case-17 | fail→fail | 13,512 | 4,386 | -68% | 1 | 1 | 0% | 2,066 | 1,158 | -44% | 0 | 0 | — |
case-18 | fail→pass | 17,832 | 3,291 | -82% | 1 | 1 | 0% | 2,554 | 843 | -67% | 0 | 0 | — |
case-19 | fail→pass | 16,439 | 5,974 | -64% | 1 | 1 | 0% | 2,467 | 1,209 | -51% | 0 | 0 | — |
case-20 | fail→pass | 11,792 | 7,374 | -37% | 1 | 1 | 0% | 1,719 | 1,691 | -2% | 0 | 0 | — |
case-21 | fail→pass | 16,428 | 2,453 | -85% | 1 | 1 | 0% | 2,534 | 819 | -68% | 0 | 0 | — |
case-22 | pass→pass | 15,841 | 6,556 | -59% | 1 | 1 | 0% | 2,437 | 1,496 | -39% | 0 | 0 | — |
case-23 | pass→pass | 17,966 | 5,962 | -67% | 1 | 1 | 0% | 2,381 | 1,408 | -41% | 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 21 counted toward the lift figure. The other 2 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 +57 percentage points is the difference between those two pass rates over the 21 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.