no-hallucination

To prevent AI hallucinations, mandate a strict honesty policy that requires the AI to systematically check codebases, configurations, and external documentation before generating any output. This approach ensures accuracy by refusing to produce speculative or unverified information.

Confidence-based output prioritizes accuracy by prompting the AI to articulate its uncertainties and ask clarifying questions when unsure. It mandates stopping the generation process to verify information against project codebases and documentation before proceeding with any response.

Verify API calls and configurations by guiding the AI to systematically search the project's codebase and external documentation before writing any code. This ensures the AI doesn't introduce errors and maintains strict accuracy regarding necessary API implementations.

AI safety requires stopping the generation process to ask for clarification whenever unsure about any part of a task. This prevents the AI from generating speculative or unverified content, thereby maintaining trust and preventing errors in factual statements.

Enforce a no hallucination policy when implementing complex features, writing configurations, or generating factual statements to prevent bugs and system failures. It is critical whenever maintaining trust and preventing errors in AI-generated outputs is required.

Limitations include the reliance on the AI to correctly articulate its own uncertainties and the prerequisite of manually verifying external documentation. The policy cannot guarantee safety if the AI fails to recognize its own lack of confidence during generation.