Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Optimize Finta fundraise workflow efficiency. Trigger with phrases like "finta performance", "finta efficiency", "optimize finta".
.claude/skills/jeremylongshore-finta-performance-tuning/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 54% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -36% | 0% |
Finta's fundraising API handles investor list pagination, round data aggregation, and CRM sync batching. Founders querying large investor databases (1,000+ contacts) hit pagination bottlenecks, while round aggregation across multiple funding stages compounds latency. Optimizing paginated fetches with cursor-based iteration, caching investor profiles, and batching CRM sync writes reduces pipeline load times by 50-70% and keeps fundraising dashboards responsive during active rounds.
typescriptconst cache = new Map<string, { data: any; expiry: number }>(); const TTL = { investors: 600_000, rounds: 300_000, pipeline: 120_000 }; async function cached(key: string, ttlKey: keyof typeof TTL, fn: () => Promise<any>) { const entry = cache.get(key); if (entry && entry.expiry > Date.now()) return entry.data; const data = await fn(); cache.set(key, { data, expiry: Date.now() + TTL[ttlKey] }); return data; } // Investor profiles change rarely (10 min). Pipeline stages are volatile (2 min).
typescriptasync function syncInvestorsBatch(client: any, cursor?: string, pageSize = 100) { const allInvestors = []; let nextCursor = cursor; do { const page = await client.listInvestors({ cursor: nextCursor, limit: pageSize }); allInvestors.push(...page.data); nextCursor = page.next_cursor; if (nextCursor) await new Promise(r => setTimeout(r, 200)); } while (nextCursor); return allInvestors; }
typescriptimport { Agent } from 'https'; const agent = new Agent({ keepAlive: true, maxSockets: 8, maxFreeSockets: 4, timeout: 30_000 }); // Finta API calls are lightweight — moderate socket count suffices
typescriptasync function withRateLimit(fn: () => Promise<any>): Promise<any> { const res = await fn(); const remaining = parseInt(res.headers?.['x-ratelimit-remaining'] || '50'); if (remaining < 3) { const resetMs = parseInt(res.headers?.['x-ratelimit-reset'] || '5') * 1000; await new Promise(r => setTimeout(r, resetMs)); } return res; }
typescriptconst metrics = { apiCalls: 0, cacheHits: 0, syncErrors: 0, avgLatencyMs: 0 }; function track(startMs: number, cached: boolean, error?: boolean) { metrics.apiCalls++; metrics.avgLatencyMs = (metrics.avgLatencyMs * (metrics.apiCalls - 1) + (Date.now() - startMs)) / metrics.apiCalls; if (cached) metrics.cacheHits++; if (error) metrics.syncErrors++; }
| Issue | Cause | Fix | |-------|-------|-----| | Slow investor list load | Offset-based pagination on large dataset | Switch to cursor-based iteration with limit=100 | | Stale round totals | Aggregation cache too long during active round | Reduce round TTL to 5 min, invalidate on write | | CRM sync timeout | Too many individual writes | Batch CRM updates in groups of 50 | | 429 Rate Limited | Burst of API calls during pipeline refresh | Parse rate limit headers, add progressive backoff |
See finta-reference-architecture.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 36,723 | 37,689 | +3% | 1 | 1 | 0% | 8,256 | 8,424 | +2% | 0 | 0 | — |
case-02 | fail→fail | 29,091 | 23,806 | -18% | 1 | 1 | 0% | 4,890 | 5,665 | +16% | 0 | 0 | — |
case-03 | fail→pass | 35,224 | 19,501 | -45% | 1 | 1 | 0% | 6,596 | 5,519 | -16% | 0 | 0 | — |
case-04 | fail→fail | 24,333 | 19,649 | -19% | 1 | 1 | 0% | 3,443 | 3,786 | +10% | 0 | 0 | — |
case-05 | fail→fail | 33,817 | 12,196 | -64% | 1 | 1 | 0% | 1,194 | 2,365 | +98% | 0 | 0 | — |
case-06 | pass→pass | 14,381 | 16,787 | +17% | 1 | 1 | 0% | 2,395 | 3,014 | +26% | 0 | 0 | — |
case-07 | fail→pass | 10,351 | 4,568 | -56% | 1 | 1 | 0% | 1,709 | 1,800 | +5% | 0 | 0 | — |
case-08 | fail→pass | 6,961 | 4,088 | -41% | 1 | 1 | 0% | 1,145 | 1,762 | +54% | 0 | 0 | — |
case-09 | fail→pass | 14,745 | 11,646 | -21% | 1 | 1 | 0% | 1,946 | 2,284 | +17% | 0 | 0 | — |
case-10 | fail→pass | 13,363 | 7,845 | -41% | 1 | 1 | 0% | 2,436 | 1,550 | -36% | 0 | 0 | — |
case-11 | pass→pass | 19,873 | 4,401 | -78% | 1 | 1 | 0% | 2,675 | 1,769 | -34% | 0 | 0 | — |
case-12 | fail→pass | 15,665 | 3,450 | -78% | 1 | 1 | 0% | 1,905 | 1,580 | -17% | 0 | 0 | — |
case-13 | pass→pass | 20,579 | 12,408 | -40% | 1 | 1 | 0% | 2,634 | 2,377 | -10% | 0 | 0 | — |
case-14 | fail→pass | 11,888 | 4,021 | -66% | 1 | 1 | 0% | 1,970 | 1,795 | -9% | 0 | 0 | — |
case-15 | fail→pass | 19,522 | 7,443 | -62% | 1 | 1 | 0% | 2,436 | 2,348 | -4% | 0 | 0 | — |
case-16 | fail→pass | 16,494 | 4,170 | -75% | 1 | 1 | 0% | 2,060 | 1,898 | -8% | 0 | 0 | — |
case-17 | pass→pass | 14,334 | 7,204 | -50% | 1 | 1 | 0% | 1,859 | 2,424 | +30% | 0 | 0 | — |
case-18 | fail→fail | 20,988 | 10,515 | -50% | 1 | 1 | 0% | 3,137 | 3,178 | +1% | 0 | 0 | — |
case-19 | pass→pass | 19,140 | 17,878 | -7% | 1 | 1 | 0% | 2,596 | 3,623 | +40% | 0 | 0 | — |
case-20 | pass→pass | 8,734 | 7,638 | -13% | 1 | 1 | 0% | 1,437 | 1,434 | -0% | 0 | 0 | — |
case-21 | fail→pass | 14,726 | 10,405 | -29% | 1 | 1 | 0% | 2,602 | 2,080 | -20% | 0 | 0 | — |
case-22 | fail→pass | 9,252 | 9,951 | +8% | 1 | 1 | 0% | 1,701 | 2,060 | +21% | 0 | 0 | — |
case-23 | fail→pass | 6,605 | 2,904 | -56% | 1 | 1 | 0% | 1,289 | 1,549 | +20% | 0 | 0 | — |
case-24 | fail→pass | 17,923 | 10,378 | -42% | 1 | 1 | 0% | 2,726 | 3,075 | +13% | 0 | 0 | — |
case-25 | fail→fail | 15,185 | 11,659 | -23% | 1 | 1 | 0% | 2,994 | 3,495 | +17% | 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. 25 cases were attempted, and 24 counted toward the lift figure. The other 1 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 +52 percentage points is the difference between those two pass rates over the 24 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.
Other measured skills in the registry, with their headline benchmark lift.