auditing-prompt-token-budget

What problem does it solve?

It helps you find why an Anthropic API prompt is getting expensive, slow, or close to the context window by measuring token usage, identifying stable versus volatile content, and exposing cache opportunities.

Core Features & Use Cases

• Exact prompt budgeting: Counts system text, tool definitions, messages, and tool results so you can trust the numbers behind your cost audit.
• Cache placement guidance: Shows where to put cache_control on the last block of the stable region so the right prefix is cached.
• TTL selection: Recommends 5m for short in-conversation reuse or 1h for cross-call reuse across hours or days.
• Cost and risk analysis: Compares write cost, read cost, and no-cache cost while flagging bad patterns like duplicated boilerplate, volatile data in cached blocks, or segments that are too small to cache.
• Operational monitoring: Encourages telemetry for cache creation and cache reads so teams can track hit rate and catch regressions early.

Quick Start

Ask this skill to audit your Anthropic prompt, identify the stable prefix, recommend the correct cache_control boundary, and choose the right TTL for your reuse pattern.