ML engineers responsible for system prompt maintenance find that prompts grow to thousands of tokens over time as new rules and examples are appended, increasing cost and degrading instruction-following due to context dilution. This agent audits, compresses, and restructures bloated system prompts without losing behavioral coverage.