ai-agent-cost-optimizerlisted

# AI Agent Cost Optimizer Reduce AI agent spend without reducing shipped quality. Optimize for cost per correct completed task, not raw token count. ## When to Activate Activate this skill when: - AI coding bills, token usage, API inference cost, or model spend are too high - Reviewing agent workflows for context waste, retry loops, or cache misses - Choosing model tiers for planning, implementation, review, or cleanup tasks - Evaluating claims about prompt caching, cheap models, or token savings - Capturing repeated workflows so future agents do not rediscover the same context ## Core Principle Price the model and context to the cost of failure. Use more capable models when a wrong answer could create expensive rework, security risk, data loss, or architectural drag. Use cheaper models and smaller context when the task is bounded, reversible, test-covered, or mechanical. Do not save tokens in ways that increase retries, hide important evidence, or lower code quality. ## Cost Audit Workflow ### 1. Establish the Spend Shape Identify what is actually driving cost before recommending changes: - API versus subscription spend - Providers and models in use - Workflows that run most often - Average input, output, cached input, reasoning, and tool-call tokens when available - Retry count, failed runs, and human correction time - Whether costs come from a few large workflows or many small calls If usage data is unavailable, mark it unknown and inspect local configs, logs,