Accelerate

When your agentic workflow is burning tokens or taking too long to respond, this walks you through a systematic performance audit. It measures baseline metrics like p50/p95 latency and cost per request, then guides you through specific optimizations: token reduction, model cascading to cheaper models for simple tasks, caching strategies, and parallelization. The structure is methodical, maybe overly so, but that's the point. It forces you to measure before and after each change and verify quality didn't regress with golden tests. The trade-off documentation is smart because speed improvements always cost something, even if it's just added complexity. Best used when you've got a working workflow that's too expensive to run at scale.

skills.sh installs: 344
GitHub stars: 602
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add sharpdeveye/maestro --skill accelerate --agent claude-code

Installs into .claude/skills of the current project.

Files

MANDATORY PREPARATION

Invoke /agent-workflow — it contains workflow principles, anti-patterns, and the Context Gathering Protocol. Follow the protocol before proceeding — if no workflow context exists yet, you MUST run /teach-maestro first. Consult the context-management reference in the agent-workflow skill for window optimization and budget strategies.

Make the workflow faster and cheaper without sacrificing quality. Measure before and after.

Performance Audit

Measure current performance:

Current metrics:
  Latency (p50): ___ms
  Latency (p95): ___ms
  Cost per request: $___
  Token usage (avg): ___ input / ___ output
  Error rate: ___%