Jonathon's AI Wiki

LinkedIn Thought Leadership Prompts (God of Prompt)

5 min read

Source: Godofprompt Amy Edmondson Linkedin Thought Leadership Prompts — thread by @godofprompt on X, extracted via browser automation 2026-04-12.

A five-prompt Claude workflow for building LinkedIn thought leadership: narrative arc, signature idea, 6-week content plan, proof stack, and comment-as-leads strategy. Each prompt produces one concrete deliverable the user can run that week. Uses the same five-slot skeleton as the author’s Winston and Godin threads but expressed with markdown headers (#ROLE:, #TASK:, #STEPS:, #RULES:, #OUTPUT:) instead of XML tags — a useful data point for when to pick which syntax.

Key Takeaways

  • Markdown headers vs XML tags: same skeleton, different syntax. Compare with Winston and Godin which use <role>/<task>/<steps>/<rules>/<output>. The five-slot pattern is the constant. Anthropic’s docs prefer XML for parseable structure; markdown headers are more readable when a human will paste the prompt manually into claude.ai.
  • Negative constraints are unusually specific. “No tips posts, no motivational content” / “No humble-bragging framed as vulnerability” / “No ‘great point’ openers — ever” / “Listicles only if every item is non-obvious.” These ban the exact weak LinkedIn patterns the user is likely to default to.
  • Every #OUTPUT: line is a directed-graph pipeline, e.g., Belief Scores → Strongest Belief → Named Framework → 2-Sentence Explanation → Counterintuitive Core. Predictable deliverable shape without requiring JSON schema.
  • Prompts 1-4 produce content; prompt 5 produces distribution strategy. The thread isn’t just “write posts” — it structures a full LinkedIn authority flywheel (narrative → positioning → calendar → proof → visibility).
  • The “Amy Edmondson” framing is aspirational, not methodological. The prompts don’t cite or apply Edmondson’s psychological safety or “teaming” frameworks. Treat the attribution as hook, not substance. ^[inferred]
  • Step 1 of every prompt is “ask first.” Same interview-pattern default the author uses in all their threads — non-optional.

The Five Prompts

Full prompt text in the source: Godofprompt Amy Edmondson Linkedin Thought Leadership Prompts. Summary:

  1. Build Your Leadership Story — 3-chapter narrative (who I was → turning point → what I now see), each chapter becomes a content pillar.
  2. Find Your Signature Idea — score five beliefs on Originality/Specificity/Defensibility, build the strongest into a named framework with 2-sentence explanation.
  3. 6-Week LinkedIn Authority — weekly content plan using only Teaching / Challenge / Decision post types. No tips. No motivation. Every post has tension.
  4. Build Your Proof Stack — mine 3 years of work for 3 expertise moments + 1 failure + 1 profile-worthy result, each as a post brief (context → tension → insight → takeaway).
  5. Turn Comments Into Leads — 5 post types where a strong comment stands out, each with a template (data point → challenge → open question).

Pattern: Markdown-Header Prompts

When your prompt will be pasted into a chat UI by a human, markdown headers (#ROLE:, #TASK:, etc.) read cleaner than XML tags — no syntax noise, follows chat-UI rendering. XML tags are better when:

  • The prompt is embedded in a larger programmatic workflow
  • You want Claude to parse or reference sections by tag name (<steps>)
  • You’re composing a system prompt with clear structural boundaries

The pattern generalizes: pick markdown headers for ad-hoc single-shot prompts, XML tags for production prompts that ship in code or get versioned. See Prompt Engineering Essentials for the explicit-criteria and interview patterns this thread applies.

Cross-Thread Comparison

AspectWinston (Presentations)Godin (Personal Brand)Edmondson (LinkedIn)
Prompt count665
SyntaxXML tagsXML tagsMarkdown headers
Combinator prompt?NoYes (#6)No
Attribution fidelityHigh (Winston’s real framework)Medium (Godin concepts + “Star” mnemonic debated)Low (Edmondson framing is stylistic only)
Domain focusSingle talkBrand identityDistribution system

Try It

  • Run prompts 1-4 as a weekend authority sprint — produces narrative arc, framework, 6-week calendar, and post briefs in one session. Prompt 5 is distribution, run it after you’ve shipped 2-3 posts and have something to point at.
  • Port the markdown-header format to your own repeatable prompts — if you currently have an XML-tag prompt you paste manually into claude.ai, rewrite it in the #ROLE: / #TASK: / #STEPS: / #RULES: / #OUTPUT: style and see if it reads cleaner.
  • Extract the banned-pattern lists — “no tips posts,” “no humble-bragging framed as vulnerability,” “no ‘great point’ openers” — these are reusable negative constraints for any marketing-content prompt. Consider a shared banned-patterns include file.
  • Replace “Amy Edmondson” with an actual named framework — if you want the attribution to carry weight, swap in a real methodology (Andy Raskin’s strategic narrative, April Dunford’s positioning, or Edmondson’s actual “teaming”/psych-safety work) and adjust the prompts accordingly.

Open Questions

  • Why does this thread use markdown headers while the Winston and Godin threads use XML tags? Is the choice domain-driven (LinkedIn content = more human-paste, less programmatic) or just a stylistic shift over time?
  • Does the directed-graph #OUTPUT: pattern (A → B → C → D) improve output reliability vs a free-form “deliver these sections” instruction? Worth A/B testing.
  • Would replacing the Amy Edmondson frame with a real named framework (e.g., Andy Raskin’s “strategic narrative”) produce measurably stronger outputs, or is the framing purely psychological?
  • The “5 post types where comments are weak” in prompt 5 is under-specified — which post types does this target in practice? A named list (e.g., “executive thought leadership posts,” “tool recommendations,” “career milestone announcements”) would make prompt 5 more portable.

Graph View

Backlinks