Agent skill · marketing growth · jeremylongshore
salesloft-performance-tuning
Analyze and improve measured Salesloft sync latency and throughput with cursor polling, bounded page size, safe concurrency, caching, and correctness checks. Use when an integration is slow or falling behind. Trigger with "Salesloft performance", "speed up Salesloft sync", or "Salesloft cursor poller".
Why this skill is useful
Adds executable scripts for optimizing SalesLoft API performance through caching, pagination strategies, and connection pooling that aren't in public docs.
What it needs
Requires axios, lru-cache installed locally. Requires salesloft account access. About 2k tokens when loaded. Last updated 2026-08-07. 2,604 stars on the source repository.
What this skill does
Salesloft Sync Performance Tuning Overview This skill optimizes one measured Salesloft path without trading away completeness or tenant isolation. It treats deep pagination and uncontrolled concurrency as cost and correctness risks. Prerequisites A named operation, team, dataset size, and service-level objective Baseline latency, pages, endpoint cost, remaining budget, retries, and lag Correctness assertions and replayable fixtures Durable cursor and destination transaction boundary Tool Discipline Use Read, Glob, and Grep to locate pagers, caches, concurrency, checkpoints, and metrics. Use WebFetch only for current official Salesloft contracts. Use Write or Edit after the baseline and target path are confirmed. Current Contract Supported list endpoints generally page from 1 with up to 100 records per page. Deep page indices incur higher documented cost, so full scans degrade shared team capacity. Salesloft's efficient polling pattern sorts updatedat ascending and persists a microsecond-precision cursor. A production poller must handle overlap and duplicate processing safely. Actual response headers, not assumed request counts, measure endpoint cost. Authentication Use the existing tenant-bound read credential. Performance work must not broaden scopes or mix caches, cursors, or limiters between teams. Instructions 1. Measure the slow path with record count, page count, p50/p95 latency, endpoint cost, remaining budget, and lag. 2. Confirm the endpoint's supported filters and sort fields. 3. Replace repeated deep scans with an updatedat cursor, ascending order, page size up to 100, and a deliberate overlap window. 4. Persist destination data and cursor atomically; deduplicate by stable resource ID and content fingerprint. 5. Add bounded per-team concurrency and cache only data with explicit freshness rules. 6. Load-test with fixtures, then run a bounded canary against an approved team. 7. Compare before/after performance and reconciliation counts before rollout. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills jeremylongshore/salesloft-performance-tuning