Agent skill · software engineering · jeremylongshore
mindtickle-performance-tuning
Diagnose and improve a Mindtickle integration with measured latency, backlog, freshness, and reconciliation evidence. Use when reports or syncs miss objectives. Trigger with "tune Mindtickle performance".
Why this skill is useful
Adds executable scripts for optimizing MindTickle API performance through caching and batch processing that aren't in public docs.
What it needs
Requires node, redis installed locally. Requires mindtickle account access. About 2k tokens when loaded. Last updated 2026-08-07. 2,604 stars on the source repository.
What this skill does
Evidence-Driven Mindtickle Integration Performance Overview Improve the customer-controlled integration path from a reproducible baseline while respecting tenant contracts, privacy, freshness, and correctness. Prerequisites A service-level objective, representative workload, current contract digest, and observation window Metrics for latency, attempts, errors, retries, backlog age, throughput, freshness, and reconciliation A non-production load path or approved canary with explicit abort thresholds Tool Discipline Use Read, Glob, and Grep to inspect traces and configuration, WebFetch for current authorized contracts, and Write or Edit for benchmarks, controlled changes, tests, and redacted receipts. Current Contract Performance spans the customer application, network, identity, adapter, tenant operations, and downstream systems. Mindtickle states support for scalable integrations but does not publish universal latency or quota targets, so customer objectives and tenant-specific evidence must remain separate. Authentication Benchmark with synthetic or aggregate data and a least-privilege non-production principal. Never use extra credentials to create artificial concurrency or include payload contents in traces. Instructions 1. Define the user or business symptom, objective, correctness invariant, workload, and baseline window. 2. Break elapsed time into queue, customer processing, network, tenant response, downstream processing, and reconciliation. 3. Segment by operation, page or batch, payload class, tenant, response class, and time without exposing personal data. 4. Rank bottlenecks by measured contribution and distinguish latency from throttling, backlog, and data-quality delay. 5. Propose one controlled change at a time: scheduling, bounded concurrency, pagination, batching, caching, indexing, or payload reduction only where the contract permits it. 6. Define cache ownership, freshness, invalidation, privacy, and stale-data behavior before enabling a cache. 7. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills jeremylongshore/mindtickle-performance-tuning