Claude-optimized prompt structure shaped for chatgpt.
# Task
Draft an internal policy. Use ONLY the provided requirements and constraints.
# Inputs- Policy topic: {policy_topic}- Goal: {goal}- Audience: {audience}- Requirements: {requirements}- Constraints: {constraints}- Known exceptions: {known_exceptions}# Anti-hallucination rules1. Do not invent legal requirements, regulations, or compliance standards.
2. If a rule needs legal review, mark it [VERIFY WITH LEGAL].
3. Do not assume organizational structure — use roles from the inputs.
# Policy structure1. **Header**: Policy name, version (Draft 1.0), effective date (TBD), owner, review cycle (suggested).
2. **Purpose** (2-3 sentences): Why this policy exists. Written so anyone can understand the intent.
3. **Scope**: Who and what is covered. What's explicitly excluded.
4. **Definitions**: Key terms used in the policy (if any need clarification).
5. **Policy rules**: Numbered, unambiguous statements. Each rule must pass the enforceability test:
- Is it clear enough that two people would interpret it the same way?
- Is it measurable or observable?
- Is it within the organization's control?
Flag any rule that fails with [NEEDS CLARIFICATION].
6. **Roles and responsibilities**: Who enforces, who complies, who approves exceptions.
7. **Exceptions**: Under what conditions can the policy be overridden? Who approves?
8. **Violations and consequences**: What happens when the policy is broken? Graduated response (warning → escalation → formal action).
9. **Review and updates**: How often is the policy reviewed? What triggers an out-of-cycle review?
10. **Related policies**: Cross-references to adjacent policies (if mentioned in inputs).
# Writing rules- Plain language. No legalese unless the topic requires it.
- Active voice. "Employees must..." not "It is required that..."
- One rule per numbered item.
# Output
Return the complete policy draft using all 10 sections.
claudegrounded_facts
claude variant
Claude-optimized prompt structure shaped for claude.
<context><policy_topic>{policy_topic}</policy_topic><goal>{goal}</goal><audience>{audience}</audience><requirements>{requirements}</requirements><constraints>{constraints}</constraints><known_exceptions>{known_exceptions}</known_exceptions></context><task>Draft an internal policy using ONLY provided requirements.</task><instructions>
No invented legal requirements. Mark uncertain rules [VERIFY WITH LEGAL]. Use input roles only.
Structure: Header (name, version, owner, review cycle) > Purpose > Scope > Definitions > Policy rules (numbered, enforceable — flag unclear ones [NEEDS CLARIFICATION]) > Roles > Exceptions > Violations (graduated) > Review cadence > Related policies.
Plain language, active voice, one rule per item.
</instructions><output>
Return the complete policy with all 10 sections.
</output>
geminigrounded_facts
gemini variant
Claude-optimized prompt structure shaped for gemini.
Policy topic: {policy_topic}
Goal: {goal}
Audience: {audience}Requirements:{requirements}Constraints:{constraints}
Known exceptions: {known_exceptions}
Rules: only provided info, mark legal gaps [VERIFY WITH LEGAL], flag unclear rules [NEEDS CLARIFICATION]. Plain language, active voice.
Structure: Header > Purpose > Scope > Definitions > Numbered rules (enforceable) > Roles > Exceptions > Violations (graduated) > Review cadence > Related policies.
Based on the entire content above, return the complete policy draft.
Related templates
More templates in the same working category.
Meeting Summary
Summarize meetings with structured decisions, action items, disagreement tracking, and confidence ratings.