Brand Voice and Safety Guardrail Audit for Automated Copywriting

Role: Principal Brand Governance Lead specializing in generative marketing compliance and voice preservation.

Context

• Brand name: {{brand_name}}
• Voice guidelines and tone boundaries: {{brand_voice_guidelines}}
• Prohibited topics, banned terms, and claim restrictions: {{prohibited_claim_topics}}
• Active distribution channels: {{marketing_channels}}
• Scope of autonomous copy generation: {{ai_copywriter_scope}}
• Target customer segment: {{target_audience}}

Task

Formulate a rigorous Brand Safety and Tone Guardrail Audit Report for {{brand_name}} that establishes deterministic system boundaries, defensive negative prompt rules, and automated human-in-the-loop escalation triggers across all {{marketing_channels}}.

Method

1. Evaluate {{ai_copywriter_scope}} to identify high-risk generative exposure points across {{marketing_channels}}.
2. Translate {{brand_voice_guidelines}} into measurable, binary guardrail metrics for tone, sentiment, and reading grade level tailored to {{target_audience}}.
3. Map all items in {{prohibited_claim_topics}} into exact-match blocklists and semantic-similarity threshold triggers.
4. Design pre-generation prompt constraints to restrict hallucinations and hyperbole in value proposition statements.
5. Construct post-generation automated heuristic filters to inspect outputs for banned terms and aggressive competitive disparagement.
6. Formulate deterministic routing logic specifying when the system halts output and routes copy to a human editor.
7. Detail a continuous validation test suite with adversarial prompt injection and edge-case testing scenarios.

Constraints

• MUST express all tone boundaries as quantifiable check criteria rather than subjective descriptions.
• MUST NOT permit any autonomous bypass for high-visibility channels listed in {{marketing_channels}}.
• Every prohibited claim in {{prohibited_claim_topics}} MUST have an explicit programmatic mitigation step.
• Limit recommendations to non-intrusive runtime checks that preserve low generation latency.

Output format

Provide a formal four-section report:

1. Executive Threat & Safety Overview (150 words max)
2. Deterministic Guardrail Architecture (table: Rule Name, Trigger Condition, Enforcement Action, Severity Level)
3. Negative Prompting & System Injection Specifications (exact system prompt directives)
4. Human-in-the-Loop Escalation & Recovery Protocol (step-by-step workflow with fallback templates)

Self-review

• Confirm all items in {{prohibited_claim_topics}} are mapped to specific enforcement actions.
• Verify that every channel in {{marketing_channels}} has a tailored risk tolerance score.
• Ensure the prompt directives contain direct, testable negative constraints without ambiguity.