Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Diagnose ClickHouse runtime query failures when the user wants database-level cause and fix guidance from an error or numeric error code, not source-code root cause analysis.
.claude/skills/frankchen021-diagnose-clickhouse-errors/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✓→✗ | ▼ Worse | 35% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -63% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 32% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | -36% | 0% |
| case-01 | ✗→✗ | ＝ Same ✗ | -42% | 0% |
Do not use this skill when the user's primary goal is:
In those cases, prefer source-code-inspection. Do not start this skill's error-code intake flow unless the user is asking for database-level diagnosis.
source-code-inspection.Code: 60).ask_user_question with exactly one question. Do NOT reply with natural-language text before the tool call.header: Please provide a ClickHouse error code for diagnosisoptions: [ { "id": "error_code", "label": "error code", "input": "text" } ]value as the numeric error code and continue.references/<code>.md with skill_resource (e.g. references/60.md) and follow its workflow.skill_resource returns nothing, use your ClickHouse knowledge to provide a best-effort response.Heading rules:
## Cause, ## Fix, and optional ## Example.Response language (BCP-47): …), localize the heading text to that language while keeping the same 3-section structure.Keep answers brief and action-first. Do not repeat the raw error verbatim. Do not add extra headings.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,611 | 7,058 | -7% | 1 | 1 | 0% | 1,309 | 754 | -42% | 0 | 0 | — |
case-02 | fail→fail | 11,174 | 4,335 | -61% | 1 | 1 | 0% | 1,899 | 689 | -64% | 0 | 0 | — |
case-03 | fail→fail | 10,381 | 4,605 | -56% | 1 | 1 | 0% | 1,767 | 739 | -58% | 0 | 0 | — |
case-04 | pass→fail | 5,218 | 2,710 | -48% | 1 | 1 | 0% | 706 | 952 | +35% | 0 | 0 | — |
case-05 | pass→pass | 5,588 | 2,848 | -49% | 1 | 1 | 0% | 732 | 966 | +32% | 0 | 0 | — |
case-06 | pass→pass | 13,225 | 7,221 | -45% | 1 | 1 | 0% | 2,289 | 1,459 | -36% | 0 | 0 | — |
case-07 | fail→fail | 9,503 | 5,573 | -41% | 1 | 1 | 0% | 1,516 | 794 | -48% | 0 | 0 | — |
case-08 | fail→fail | 4,337 | 4,190 | -3% | 1 | 1 | 0% | 849 | 773 | -9% | 0 | 0 | — |
case-09 | fail→fail | 12,836 | 5,892 | -54% | 1 | 1 | 0% | 2,059 | 714 | -65% | 0 | 0 | — |
case-10 | fail→fail | 15,615 | 4,041 | -74% | 1 | 1 | 0% | 2,779 | 724 | -74% | 0 | 0 | — |
case-11 | fail→fail | 8,468 | 4,963 | -41% | 1 | 1 | 0% | 1,458 | 770 | -47% | 0 | 0 | — |
case-12 | fail→fail | 11,564 | 4,529 | -61% | 1 | 1 | 0% | 1,870 | 767 | -59% | 0 | 0 | — |
case-13 | fail→fail | 12,229 | 3,966 | -68% | 1 | 1 | 0% | 1,963 | 714 | -64% | 0 | 0 | — |
case-14 | fail→fail | 12,311 | 4,606 | -63% | 1 | 1 | 0% | 2,237 | 758 | -66% | 0 | 0 | — |
case-15 | fail→fail | 15,909 | 6,045 | -62% | 1 | 1 | 0% | 2,722 | 758 | -72% | 0 | 0 | — |
case-16 | fail→fail | 15,591 | 4,656 | -70% | 1 | 1 | 0% | 2,598 | 721 | -72% | 0 | 0 | — |
case-17 | fail→fail | 11,805 | 5,318 | -55% | 1 | 1 | 0% | 1,925 | 696 | -64% | 0 | 0 | — |
case-18 | fail→fail | 14,638 | 6,397 | -56% | 1 | 1 | 0% | 2,547 | 731 | -71% | 0 | 0 | — |
case-19 | fail→fail | 13,372 | 4,069 | -70% | 1 | 1 | 0% | 2,279 | 693 | -70% | 0 | 0 | — |
case-20 | fail→fail | 14,921 | 4,308 | -71% | 1 | 1 | 0% | 2,380 | 699 | -71% | 0 | 0 | — |
case-21 | pass→fail | 11,259 | 8,985 | -20% | 1 | 1 | 0% | 2,013 | 744 | -63% | 0 | 0 | — |
case-22 | fail→fail | 22,480 | 6,880 | -69% | 1 | 1 | 0% | 1,834 | 871 | -53% | 0 | 0 | — |
case-23 | fail→fail | 14,164 | 5,467 | -61% | 1 | 1 | 0% | 2,207 | 724 | -67% | 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 3 counted toward the lift figure. The other 20 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 -100 percentage points is the difference between those two pass rates over the 3 comparable cases. 11 cases got worse with the skill loaded, and they are included in that figure.
Other measured skills in the registry, with their headline benchmark lift.