# Auditing the vocabulary of system prompts and instruction files > Source-anchored research for Auditing the vocabulary of system prompts and instruction files. Run status: COMPLETED. Parent: [Prompt-side control of LLM lexical tics](https://llms-explorer.com/tree/ai-isms-prompt-side-control/) · 3 facets · 5 facts · page: https://llms-explorer.com/tree/auditing-the-vocabulary-of-system-prompts-and-instruction-files/ ## How it works - Anthropic advises that on Opus 4.5/4.6-era models, aggressive emphasis such as 'CRITICAL: You MUST use this tool when...' causes overtriggering; replace with plain 'Use this tool when...'. Audit step: strip shouted/emphatic vocabulary from prompts written for older models. (medium confidence) (verified as of 2026-09-23) — [source](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices) ## Comparisons and alternatives - Lexical overuse in LLM output (e.g. 'delve', 'nuanced') is traced to human-feedback training, not prompts: Llama Instruct vs Base shows large increases ('nuanced' +8,342%), and human raters preferred text with these words (52.4% vs 47.6%). A prompt audit cannot remove the underlying bias. (medium confidence) — [source](https://arxiv.org/html/2508.01930v1) ## Facts and statements - Anthropic docs state prompt formatting style can influence response style; removing markdown from the prompt can reduce markdown in output. Prompt vocabulary and format are therefore an audit surface. (low confidence) (verified as of 2026-09-23) — [source](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices) - Published system-prompt word-ban lists exist (e.g. LLM-Reset lists 'crucial', 'delve', 'matters'); they are practitioner artifacts and their effect is not validated in the papers surveyed. (low confidence) — [source](https://lifearchitect.ai/llm-reset/) - Banning a word or phrase in a prompt can suppress a whole cluster of related constructions, so an audit that adds ban lists risks side effects. Source is a practitioner essay, not a controlled study. (low confidence) — [source](https://habr.com/en/articles/1032620/)