Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Query CAST AI cluster savings report and node inventory. Use when verifying CAST AI connectivity, viewing cluster cost savings, or listing managed nodes after onboarding. Trigger with phrases like "cast ai hello world", "cast ai savings", "cast ai cluster status", "test cast ai connection".
.claude/skills/jeremylongshore-castai-hello-world/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 103% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 36% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 78% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -36% | 0% |
Prove the minimum useful loop: identify one sandbox cluster, preview the connection, connect with Cost Monitoring, verify telemetry, and leave optimization automation unchanged.
Use Bash(kubectl:\) to print the current context and inspect only basic cluster identity. Use Read and Grep to ensure the cluster is not already owned by CAST AI, another autoscaler, Terraform, or GitOps.
Use Bash(castctl:\) to check the client version and run the documented cluster connection dry-run. Review detected provider, cluster name, region, organization, proposed features, namespace, and cloud changes.
Use Write to capture the dry-run, selected cluster, owners, features, start time, expected telemetry, and cleanup path. Cost Monitoring is always enabled; do not select Node or Workload Autoscaling merely to complete a quickstart.
Run the reviewed interactive connection. Confirm the prompt resolves the same cluster and organization. Keep the returned console URL, but never store the browser token, kubeconfig, or API key in the receipt.
Use Bash(kubectl:\) to verify castai-agent namespace workloads, readiness, and recent warning events. Confirm the console identifies the correct cluster and begins showing cost data. Treat an empty initial report as ingestion time, not proof of zero cost.
If the evaluation continues, hand off to observation-first onboarding. If it ends, use the documented disconnect path only after reviewing cloud and cluster cleanup effects and preserving the final receipt.
Use Read and Grep for ownership checks. Use Write for the sanitized quickstart receipt. Use Bash(castctl:_) for documented dry-run, connection, and status operations and Bash(kubectl:_) for bounded verification only.
A team connects a sandbox EKS cluster for Cost Monitoring, confirms agent readiness, and waits for representative data. It does not enable node automation or change workload requests during the first session.
| Failure | Response | | -------------------------------------------- | -------------------------------------------------------- | | The context is production | Stop and select an approved sandbox | | Dry-run detects the wrong provider or region | Correct environment selection before connecting | | Agent pods fail readiness | Collect bounded evidence and do not add features | | Cleanup consequences are unclear | Leave state unchanged and escalate to the platform owner |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 9,007 | 3,119 | -65% | 1 | 1 | 0% | 1,857 | 1,340 | -28% | 0 | 0 | — |
case-02 | fail→fail | 13,279 | 5,448 | -59% | 1 | 1 | 0% | 2,379 | 1,971 | -17% | 0 | 0 | — |
case-03 | fail→pass | 21,122 | 7,163 | -66% | 1 | 1 | 0% | 1,149 | 2,328 | +103% | 0 | 0 | — |
case-14 | pass→pass | 8,668 | 4,090 | -53% | 1 | 1 | 0% | 1,119 | 1,614 | +44% | 0 | 0 | — |
case-04 | fail→pass | 5,724 | 2,821 | -51% | 1 | 1 | 0% | 964 | 1,312 | +36% | 0 | 0 | — |
case-05 | fail→pass | 8,920 | 3,823 | -57% | 1 | 1 | 0% | 1,555 | 1,504 | -3% | 0 | 0 | — |
case-06 | pass→fail | 5,508 | 2,545 | -54% | 1 | 1 | 0% | 899 | 1,330 | +48% | 0 | 0 | — |
case-07 | fail→pass | 5,129 | 4,033 | -21% | 1 | 1 | 0% | 867 | 1,539 | +78% | 0 | 0 | — |
case-08 | fail→pass | 11,781 | 2,307 | -80% | 1 | 1 | 0% | 2,047 | 1,309 | -36% | 0 | 0 | — |
case-09 | pass→pass | 4,941 | 3,553 | -28% | 1 | 1 | 0% | 813 | 1,299 | +60% | 0 | 0 | — |
case-10 | pass→pass | 7,449 | 3,606 | -52% | 1 | 1 | 0% | 1,277 | 1,445 | +13% | 0 | 0 | — |
case-11 | pass→pass | 3,925 | 2,545 | -35% | 1 | 1 | 0% | 623 | 1,230 | +97% | 0 | 0 | — |
case-12 | pass→pass | 13,015 | 7,529 | -42% | 1 | 1 | 0% | 1,512 | 2,291 | +52% | 0 | 0 | — |
case-13 | pass→pass | 11,776 | 6,650 | -44% | 1 | 1 | 0% | 2,059 | 2,171 | +5% | 0 | 0 | — |
case-15 | pass→pass | 10,061 | 4,970 | -51% | 1 | 1 | 0% | 1,721 | 1,734 | +1% | 0 | 0 | — |
case-16 | pass→pass | 9,070 | 4,459 | -51% | 1 | 1 | 0% | 1,596 | 1,606 | +1% | 0 | 0 | — |
case-17 | fail→pass | 8,323 | 2,554 | -69% | 1 | 1 | 0% | 1,395 | 1,280 | -8% | 0 | 0 | — |
case-18 | fail→pass | 7,605 | 2,257 | -70% | 1 | 1 | 0% | 1,388 | 1,159 | -16% | 0 | 0 | — |
case-19 | pass→pass | 7,106 | 3,002 | -58% | 1 | 1 | 0% | 1,310 | 1,447 | +10% | 0 | 0 | — |
case-20 | fail→fail | 19,894 | 17,002 | -15% | 1 | 1 | 0% | 3,995 | 4,465 | +12% | 0 | 0 | — |
case-21 | fail→fail | 30,014 | 10,438 | -65% | 1 | 1 | 0% | 1,132 | 2,719 | +140% | 0 | 0 | — |
case-22 | pass→pass | 7,944 | 6,240 | -21% | 1 | 1 | 0% | 1,355 | 1,932 | +43% | 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 20 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 +27 percentage points is the difference between those two pass rates over the 20 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.