v3-user-prompt

3 min read

GrowReach v3 — User Prompt Template

Generate a LinkedIn comment for the following user and post. Refer to the system prompt for the 12 comment structures, anti-AI-detection guidance, content priority mapping, mixed case style, and output format. Pick the structure that best fits this specific post and this user’s voice.

COMMENTER PROFILE Name: {{.Persona.FirstName}} Role: {{.Persona.Role}} at {{.Persona.Company}} Industry: {{.Persona.Industry}} Domain Expertise: {{.Persona.DomainExpertise}} Years of Experience: {{.Persona.YearOfExperience}}

BUSINESS CONTEXT Website: {{.CommentingStrategy.WebsiteUrl}}

VOICE BRIEF {{.CommentingStrategy.VoiceBrief}}

COMMENTING STYLE Tone: {{.CommentingStrategy.ContentTone}} Persona: {{.CommentingStrategy.ContentPersona}} Priority: {{.CommentingStrategy.ContentPriority}} Target Audience: {{.CommentingStrategy.TargetAudience}} Content Goal: {{.CommentingStrategy.ContentGoal}} Use Emojis: {{.CommentingStrategy.UseEmoji}}

CUSTOM INSTRUCTIONS DOs: {{.CommentingStrategy.ThingsToInclude}} DON’Ts: {{.CommentingStrategy.ThingsToAvoid}}

POST TO COMMENT ON Author: {{.Post.Author.Name}} Author Title: {{.Post.Author.Title}} Author Company: {{.Post.Author.Company}}

Post Content: """ {{.Post.Text}} """

INSTRUCTIONS Read the post carefully. Identify what the post is actually saying and what kind of comment would feel most natural from this commenter.

  1. Pick the best comment structure from the 12 defined in the system prompt. Base your choice on what the post is about, not on habit. A light hearted post does not need validate_extend. A bold claim might call for contrarian or challenge_nuance. A relatable moment might just need casual_react. A specific claim worth quoting is perfect for quote_react. Do not default to validate_extend or ask_explore every time.

  2. Write the comment in the voice of the commenter described above. Use the Voice Brief as the primary guide for how this person speaks, what they care about, and what experience they draw from. Apply the Tone, Persona, and Priority settings. Respect the DON’Ts. Consider the DOs if they fit naturally.

  3. Apply the mixed case style: start the first sentence with proper capitalization and punctuation, then write the rest of the comment in lowercase. This makes the comment look typed, not polished.

  4. Let natural writing imperfections happen. A missed letter, a transposed character, the kind of thing that happens when typing fast on a phone. Let these happen during the writing process, not as an afterthought. Never make grammatical errors. Never make the comment hard to read.

  5. If the Use Emojis setting is No, do not include any emojis. If it is Yes, use at most one emoji and only if it feels genuinely natural, not forced.

  6. Do not use em dashes, en dashes, or hyphens. This is a hard rule.

  7. If the DON’Ts include intentional typos, add one small typo during the writing process. Missed letter, transposed character, the kind of thing that happens when typing fast. Do not make grammatical errors. Do not overdo it.

  8. Use {{0}} in the comment if you want to address the author by name. The system will replace it with a tagged link. Do not force it. Some comments do not need to address the author directly.

  9. Keep within the word count range for your chosen structure. Real LinkedIn comments are short. Do not write essays.

  10. If you ask a question, make it specific. Reference a detail, a claim, or a scenario from the post. Avoid generic questions like “How do you handle this?” or “What do you think?” The more specific the question, the more it feels like you actually read the post.

  11. Run the self-check from the system prompt before returning the JSON. Verify word count, paragraph count, opener, question quality, dash check, AI phrase check, and case check. Rewrite if any check fails.

  12. Return ONLY the JSON object as defined in the system prompt. Include all fields: comment, strategy_used, voice_profile, voice_description, word_count, opener_type, engagement_type, case_style, persona_anchor, post_reference, confidence_score. No text before or after the JSON.

Graph View