llm-agent-token-audit

To reduce token overhead, audit your LLM agent prompts and workspace injections to identify and prune bloated content. This skill guides you to quantify usage, classify waste like dead files, and safely remove unnecessary components to minimize costs.

The best way to audit workspace files for prompt bloat is to systematically quantify token usage and categorize waste types. This process identifies dead files, placeholder templates, and redundant content to prioritize safe deletions while preserving essential agent behavior.

You should run a prompt token audit when adding new skills or tools to an agent framework. Running the audit at this stage helps strip excess workspace files and optimize prompts to improve latency and reduce fixed per-call token costs.

Yes, you can safely prune unnecessary components from agent prompts without breaking functionality. The audit provides framework-ready recommendations that prioritize deletions and configuration changes to minimize prompt size while preserving essential behavior.

An LLM token audit identifies waste types such as dead files, placeholder templates, and redundant content. By classifying these unnecessary assets within workspace injections, you can systematically eliminate bloated prompts and reduce fixed per-call token overhead.

Adding new skills increases token costs because it often introduces excess workspace files and bloated prompts. Auditing the agent framework after these additions helps quantify the new prompt overhead and safely prune unnecessary content to maintain lean operations.