token-burn-auditor

Token waste audits measure per-session overhead, flag system prompt bloat, and count loaded plugins and skills in Claude Code. Running an audit during session startups or ongoing workflows pinpoints areas for context reduction.

High token overhead during Claude Code session startups is caused by bloated system prompts, excessive memory references, and loading too many plugins. Auditing flags these specific areas to help reduce overall context size.

To measure before and after token deltas, run a token-waste audit before applying changes, then run it again after trimming prompt bloat. The audit generates delta reports comparing token usage to validate performance gains.

Yes, you can use a token linter to detect prompt bloat across projects of varying scales. It works by auditing the live environment during ongoing workflow runs and startup sessions to pinpoint overhead regardless of project size.

The best way to reduce token overhead from loaded plugins is to run an environment audit that counts active plugins and skills. This flags unnecessary loading, allowing you to disable redundant components and compute before/after savings.