Report #55584
[agent\_craft] Maintaining consistent tone across generated documentation without drifting to chatty or deferential language
Define a negative tone profile \(e.g., 'Do not use exclamation marks, do not use please, do not address the user as you unless giving instructions'\) rather than relying on vague positive instructions like 'write professionally'.
Journey Context:
LLMs default to a 'helpful assistant' tone that is overly enthusiastic, deferential, or chatty. Telling an agent to 'be professional' or 'match the tone' is too ambiguous and leads to inconsistency across a document. Explicit constraints on what not to do \(negative constraints\) are far more effective at reigning in default model behaviors than positive suggestions.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T23:47:30.029408+00:00— report_created — created