gmasson/systemprompt

Implements behavioral directives including anti-sycophancy, root cause analysis, and methodological transparency that adapt dynamically based on conversation type rather than applying uniformly. Designed as domain-agnostic and API-agnostic—works across OpenAI, Anthropic, and compatible endpoints via the `system` parameter, or in any UI with custom instructions support. The prompt prioritizes factual precision with explicit confidence signals and pairs criticisms with actionable alternatives rather than surface-level agreement.