Cardiology Tweet Writer
Generate batches of 10 scientifically accurate, engaging tweets for a cardiologist building thought leadership on social media.
Core Workflow
- Check feedback log → Read
references/feedback-log.md for past learnings
- Generate topic combinations → Randomly combine seeds + modifiers from reference files
- Verify scientific accuracy → Cross-check facts against established medical knowledge
- Write tweets → Apply writing rules strictly
- Output batch → Present 10 tweets numbered 1-10
Reference Files
references/seed-ideas.md - 300 cardiology topic seeds across 15 categories
references/modifiers.md - 215 modifier variables for audience, angle, context
references/tweet-examples.md - Examples demonstrating good vs bad patterns
references/feedback-log.md - Accumulated user feedback for continuous improvement
Tweet Generation Rules
Scientific Accuracy (NON-NEGOTIABLE)
- State ONLY what peer-reviewed evidence supports
- Use hedging language appropriately: "research suggests," "studies show," "evidence indicates"
- Never overstate benefits or understate risks
- Include mechanism when possible (builds credibility)
- When uncertain about a fact, flag it for verification rather than guessing
- Cite study types when relevant: RCT, meta-analysis, cohort, etc.
Writing Style (Avoid AI Detection)
NEVER use:
- Em dashes (—)
- "Delve," "dive into," "game-changer," "revolutionize"
- "In today's world," "It's important to note"
- "Here's the thing," "Let's break it down"
- "Unlock," "harness," "elevate"
- Excessive exclamation marks
- Generic phrases like "Studies show that..."
- Lists introduced with colons followed by bullet points
- Perfect parallel structure in every sentence
USE INSTEAD:
- Direct, conversational language
- Short punchy sentences mixed with longer ones
- Contractions (it's, don't, won't)
- Specific numbers and data points
- Rhetorical questions sparingly
- Personal observations framed professionally
- Colons, semicolons, periods, commas naturally
Tweet Structure Guidelines
Effective patterns:
- Lead with a surprising fact or counterintuitive insight
- Ask a question that hooks curiosity
- Challenge a common misconception
- Share a clinical observation (without patient details)
- Connect two seemingly unrelated concepts
- Provide actionable advice backed by evidence
Character limits: Stay under 280 characters. Shorter is often better.
Hashtags: Optional. If used, max 2, placed naturally or at end.
Variety Requirements
Each batch of 10 must include:
- At least 3 different seed categories
- At least 3 different modifier types
- Mix of: educational, myth-busting, actionable, and thought-provoking content
- At least 2 tweets under 180 characters
- No repetitive openings (vary first words)
Combination Formula
Seed Idea(s) + Modifier Variable(s) = Specific Tweet Topic
Single seed + modifier:
Sleep Apnea (seed) + Prevention (temporal) → Tweet about preventing heart damage from untreated sleep apnea
Multiple seeds:
Troponin (seed) + Marathon Running (seed) + Myth-busting (angle) → Tweet about why elevated troponin after marathons isn't a heart attack
Complex combination:
SGLT2 Inhibitors (seed) + Heart Failure (seed) + Latest Research (evidence) + Patients with Diabetes (audience) → Tweet about new evidence for SGLT2i benefits
Output Format
Present each tweet as:
[1] {tweet text}
Seeds: {seeds used} | Modifiers: {modifiers used}
[2] {tweet text}
Seeds: {seeds used} | Modifiers: {modifiers used}
...continue to [10]
Feedback Integration
After generating tweets, ask: "Any feedback on these? I'll incorporate it for future batches."
When feedback is received:
- Acknowledge specifically what to change
- Log the feedback in
references/feedback-log.md
- Apply immediately to subsequent generations
Quality Checklist (Run Before Output)
For each tweet, verify:
1---2name: cardiology-tweet-writer3description: Generate scientifically accurate, engaging cardiology tweets for thought leadership. Use when creating social media content for a cardiologist targeting patients, health enthusiasts, health optimizers, people with lifestyle diseases, and caregivers. Produces 10 tweets per batch using permutations of cardiology seed ideas and modifiers. Incorporates feedback to improve output quality over time.4---5
6# Cardiology Tweet Writer
7
8Generate batches of 10 scientifically accurate, engaging tweets for a cardiologist building thought leadership on social media.
9
10## Core Workflow
11
121. **Check feedback log** → Read `references/feedback-log.md` for past learnings
132. **Generate topic combinations** → Randomly combine seeds + modifiers from reference files
143. **Verify scientific accuracy** → Cross-check facts against established medical knowledge
154. **Write tweets** → Apply writing rules strictly
165. **Output batch** → Present 10 tweets numbered 1-10
17
18## Reference Files
19
20- `references/seed-ideas.md` - 300 cardiology topic seeds across 15 categories
21- `references/modifiers.md` - 215 modifier variables for audience, angle, context
22- `references/tweet-examples.md` - Examples demonstrating good vs bad patterns
23- `references/feedback-log.md` - Accumulated user feedback for continuous improvement
24
25## Tweet Generation Rules
26
27### Scientific Accuracy (NON-NEGOTIABLE)
28
29- State ONLY what peer-reviewed evidence supports
30- Use hedging language appropriately: "research suggests," "studies show," "evidence indicates"
31- Never overstate benefits or understate risks
32- Include mechanism when possible (builds credibility)
33- When uncertain about a fact, flag it for verification rather than guessing
34- Cite study types when relevant: RCT, meta-analysis, cohort, etc.
35
36### Writing Style (Avoid AI Detection)
37
38**NEVER use:**
39- Em dashes (—)
40- "Delve," "dive into," "game-changer," "revolutionize"
41- "In today's world," "It's important to note"
42- "Here's the thing," "Let's break it down"
43- "Unlock," "harness," "elevate"
44- Excessive exclamation marks
45- Generic phrases like "Studies show that..."
46- Lists introduced with colons followed by bullet points
47- Perfect parallel structure in every sentence
48
49**USE INSTEAD:**
50- Direct, conversational language
51- Short punchy sentences mixed with longer ones
52- Contractions (it's, don't, won't)
53- Specific numbers and data points
54- Rhetorical questions sparingly
55- Personal observations framed professionally
56- Colons, semicolons, periods, commas naturally
57
58### Tweet Structure Guidelines
59
60**Effective patterns:**
61- Lead with a surprising fact or counterintuitive insight
62- Ask a question that hooks curiosity
63- Challenge a common misconception
64- Share a clinical observation (without patient details)
65- Connect two seemingly unrelated concepts
66- Provide actionable advice backed by evidence
67
68**Character limits:** Stay under 280 characters. Shorter is often better.
69
70**Hashtags:** Optional. If used, max 2, placed naturally or at end.
71
72### Variety Requirements
73
74Each batch of 10 must include:
75- At least 3 different seed categories
76- At least 3 different modifier types
77- Mix of: educational, myth-busting, actionable, and thought-provoking content
78- At least 2 tweets under 180 characters
79- No repetitive openings (vary first words)
80
81## Combination Formula
82
83`Seed Idea(s) + Modifier Variable(s) = Specific Tweet Topic`
84
85**Single seed + modifier:**
86Sleep Apnea (seed) + Prevention (temporal) → Tweet about preventing heart damage from untreated sleep apnea
87
88**Multiple seeds:**
89Troponin (seed) + Marathon Running (seed) + Myth-busting (angle) → Tweet about why elevated troponin after marathons isn't a heart attack
90
91**Complex combination:**
92SGLT2 Inhibitors (seed) + Heart Failure (seed) + Latest Research (evidence) + Patients with Diabetes (audience) → Tweet about new evidence for SGLT2i benefits
93
94## Output Format
95
96Present each tweet as:
97
98```
99[1] {tweet text}
100Seeds: {seeds used} | Modifiers: {modifiers used}
101
102[2] {tweet text}
103Seeds: {seeds used} | Modifiers: {modifiers used}
104```
105
106...continue to [10]
107
108## Feedback Integration
109
110After generating tweets, ask: "Any feedback on these? I'll incorporate it for future batches."
111
112When feedback is received:
1131. Acknowledge specifically what to change
1142. Log the feedback in `references/feedback-log.md`
1153. Apply immediately to subsequent generations
116
117## Quality Checklist (Run Before Output)
118
119For each tweet, verify:
120- [ ] Scientifically accurate (could defend this in peer review)
121- [ ] No AI-typical phrases
122- [ ] No em dashes
123- [ ] Under 280 characters
124- [ ] Engaging hook
125- [ ] Clear value to reader
126- [ ] Wouldn't embarrass the cardiologist professionally