Executive AI Communication Boundary Framework

Role: Principal AI Governance Strategist with 15 years of enterprise brand risk and algorithmic safety leadership.

Context

• Organization: {{company_name}}
• Operating Sector: {{industry_sector}}
• Autonomous Agent Function: {{agent_role}}
• Core Brand Voice Principles: {{brand_tone_guidelines}}
• Forbidden Discussion Vectors: {{prohibited_topics}}
• Human Escalation Threshold: {{escalation_threshold}}

Task

Design an operational brand alignment and safety guardrail framework for {{company_name}} that enforces tone boundaries, neutralizes prompt injection risks, and specifies deterministic escalation triggers for {{agent_role}}.

Method

1. Analyze {{industry_sector}} regulatory standards against the designated core duties of {{agent_role}}.
2. Map {{brand_tone_guidelines}} into measurable linguistic attributes including lexical choice, formality index, and perspective.
3. Categorize {{prohibited_topics}} into severe red-line violations versus soft-deflection topics.
4. Design deterministic input filters that screen user prompts for adversarial extraction, competitor comparisons, and policy bypass attempts.
5. Establish output validation checks that evaluate sentiment balance, claim certainty, and brand voice adherence before token emission.
6. Formulate fallback deflection scripts tailored to distinct breach severity tiers.
7. Define the technical routing mechanism and context payload dispatched at {{escalation_threshold}}.
8. Construct a post-incident telemetry schema to continuously monitor guardrail trigger rates.

Constraints

• MUST express all boundary conditions as explicit boolean evaluations or discrete deterministic categories.
• MUST NOT provide speculative legal advice or vague subjective tone descriptors.
• Fallback messaging MUST preserve brand authority without apologizing unnecessarily.
• The framework MUST be adaptable across text-only and multimodal interaction channels.

Output format

Provide a structured markdown framework organized under four distinct sections:

1. Input Filtering Taxonomy (table with columns: Category, Trigger Pattern, Action)
2. Output Validation Matrix (table with columns: Metric, Acceptable Range, Failure Action)
3. Tiered Deflection Playbook (three distinct breach severity tiers with exact copy templates)
4. Escalation and Telemetry Protocol (max 250 words describing routing logic and logged attributes)

Self-review

• Verify that all 6 variables from Context are logically integrated into the framework rules.
• Confirm that no generic placeholders exist within the Deflection Playbook templates.
• Check that every output table contains actionable operational logic rather than abstract advice.