Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Graceful Degradation with Helpful Messages
.claude/skills/graceful-degradation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -18% | 0% |
When optional services are unavailable, degrade gracefully with actionable fallback messages.
Check availability at the start, cache the result, and provide helpful messages that explain what's missing and how to fix it.
typescriptlet lmstudioAvailable: boolean | null = null; let lastCheck = 0; const CACHE_TTL = 60000; // 60 seconds async function checkLMStudio(): Promise<boolean> { const now = Date.now(); if (lmstudioAvailable !== null && now - lastCheck < CACHE_TTL) { return lmstudioAvailable; } try { const response = await fetch('http://localhost:1234/v1/models', { signal: AbortSignal.timeout(2000) }); lmstudioAvailable = response.ok; } catch { lmstudioAvailable = false; } lastCheck = now; return lmstudioAvailable; } // Usage if (!await checkLMStudio()) { return { result: 'continue', message: `LMStudio not available at localhost:1234. To enable Godel-Prover tactic suggestions: 1. Install LMStudio from https://lmstudio.ai/ 2. Load "Goedel-Prover-V2-8B" model 3. Start the local server on port 1234 Continuing without AI-assisted tactics...` }; }
[Service] not available at [endpoint].
To enable [feature]:
1. [Step to install/start]
2. [Configuration step if needed]
3. [Verification step]
Continuing without [degraded feature]...| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | fail→pass | 10,268 | 9,059 | -12% | 1 | 1 | 0% | 2,248 | 2,119 | -6% | 0 | 0 | — |
case-18 | fail→pass | 9,839 | 6,438 | -35% | 1 | 1 | 0% | 1,883 | 1,866 | -1% | 0 | 0 | — |
case-10 | fail→pass | 13,379 | 10,387 | -22% | 1 | 1 | 0% | 2,662 | 2,676 | +1% | 0 | 0 | — |
case-01 | fail→pass | 23,735 | 7,628 | -68% | 1 | 1 | 0% | 2,942 | 2,106 | -28% | 0 | 0 | — |
case-02 | fail→pass | 14,963 | 7,777 | -48% | 1 | 1 | 0% | 2,787 | 2,281 | -18% | 0 | 0 | — |
case-03 | fail→pass | 18,853 | 10,551 | -44% | 1 | 1 | 0% | 3,820 | 2,759 | -28% | 0 | 0 | — |
case-04 | fail→pass | 14,602 | 8,302 | -43% | 1 | 1 | 0% | 2,871 | 2,384 | -17% | 0 | 0 | — |
case-05 | pass→pass | 15,119 | 10,314 | -32% | 1 | 1 | 0% | 3,139 | 2,842 | -9% | 0 | 0 | — |
case-06 | fail→pass | 13,672 | 10,346 | -24% | 1 | 1 | 0% | 2,658 | 2,771 | +4% | 0 | 0 | — |
case-07 | fail→pass | 9,917 | 8,988 | -9% | 1 | 1 | 0% | 1,931 | 2,583 | +34% | 0 | 0 | — |
case-08 | fail→pass | 9,410 | 6,644 | -29% | 1 | 1 | 0% | 2,203 | 2,142 | -3% | 0 | 0 | — |
case-09 | fail→pass | 10,777 | 7,586 | -30% | 1 | 1 | 0% | 2,166 | 2,194 | +1% | 0 | 0 | — |
case-11 | fail→fail | 12,194 | 9,335 | -23% | 1 | 1 | 0% | 2,689 | 2,532 | -6% | 0 | 0 | — |
case-12 | fail→pass | 9,743 | 8,832 | -9% | 1 | 1 | 0% | 1,841 | 2,403 | +31% | 0 | 0 | — |
case-13 | fail→pass | 13,635 | 8,690 | -36% | 1 | 1 | 0% | 2,526 | 1,892 | -25% | 0 | 0 | — |
case-14 | fail→pass | 9,524 | 8,947 | -6% | 1 | 1 | 0% | 1,733 | 2,420 | +40% | 0 | 0 | — |
case-15 | fail→pass | 10,893 | 7,701 | -29% | 1 | 1 | 0% | 2,196 | 2,141 | -3% | 0 | 0 | — |
case-16 | fail→pass | 12,802 | 7,684 | -40% | 1 | 1 | 0% | 2,488 | 2,113 | -15% | 0 | 0 | — |
case-19 | fail→pass | 12,569 | 10,148 | -19% | 1 | 1 | 0% | 2,501 | 2,808 | +12% | 0 | 0 | — |
case-20 | pass→pass | 13,996 | 11,991 | -14% | 1 | 1 | 0% | 2,323 | 2,875 | +24% | 0 | 0 | — |
case-21 | fail→pass | 15,997 | 13,034 | -19% | 1 | 1 | 0% | 2,923 | 2,889 | -1% | 0 | 0 | — |
case-22 | pass→pass | 12,425 | 9,471 | -24% | 1 | 1 | 0% | 2,397 | 2,285 | -5% | 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. The headline lift of +82 percentage points is the difference between those two pass rates over the 22 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/29/2026 | +68% |
Other measured skills in the registry, with their headline benchmark lift.