← Vault Index
frameworks / kit-li-thought-leadership / 05-li-thought-leadership-output-skill.md
Source: frameworks/kit-li-thought-leadership/05-li-thought-leadership-output-skill.md

LinkedIn Thought Leadership Post — Output Skill

What this produces: A ready-to-publish LinkedIn post optimized for both authority and distribution. Plain text for text posts. If carousel, a brief with slide-by-slide content for design execution.

Output format: Text in a code block (for easy copy-paste) or saved as li-{{topic-slug}}-thought-leadership.md

Reference files:

Build Process

Step 0: Read Companion Files

Before proposing anything, read:

  1. voice.md — Vocabulary, cadence, tone, avoid list
  2. audience.md — Who the reader is, what they believe, what triggers action
  3. soul.md — Why this brand exists
  4. offer.md — What Advisory OS sells (posts don't sell directly, but the offer shapes what patterns matter)

If source material is provided, read it completely. Find the insight inside it — do not summarize.

Step 1: Run the Intake

Walk through all seven intake decisions from 01-li-thought-leadership-context.md:

  1. Angle — Propose 2–3 with rationale. Recommend one.
  2. Core Insight — The insight, the assumption it challenges, why it matters, and the status angle.
  3. Hook — Propose 2–3 opening lines.
  4. Arc — Section flow based on the angle.
  5. Conversation Prompt — Propose 2–3 prompts with status analysis for each.
  6. Closer — Propose 2–3 final lines.
  7. Format — Text post or carousel. Recommend based on the material.

After all seven are locked, produce the Post Lock:

POST LOCK: [Working Title]
Angle: [Angle name]
Core insight: [One sentence]
Assumption challenged: [What the audience currently believes]
Status angle: [What engaging signals about the reader]
Hook: [Opening line]
Arc: [Section flow]
Closer: [Final statement]
Conversation prompt: [Specific question/provocation]
Format: [Text-only / Carousel]

Step 1b: Create the Image Brief

After the Post Lock is confirmed, create the image brief:

  1. Apply the 2-second test — can someone with zero context understand this in 2 seconds?
  2. Choose the image concept — recognition question, reframing insight, or bold statement. The image stops the scroll. The post delivers the insight. The article provides the full experience.
  3. Write the quote card content:
  1. Check against anti-patterns — no crossing-lines data visuals, no data tables, no numbers without context, no too-small text, no novel frameworks as visuals.

Visual brand specs are in 02-li-thought-leadership-terminology.md (Image System section).

Step 2: Write the First Pass

Before writing a single line:

  1. Select the golden example closest to your angle — Examples 1–3 are Worldview/Insight posts, Example 4 is a Proof/Math post
  2. Study its structural spine: How does it open? Where does the story live? When does the insight land? Is it scene-led or thesis-led?
  3. Write from that structure. The Post Lock tells you WHAT to write. The golden example tells you HOW.

Two post types (Steve's DNA): All Advisory OS TL posts follow the same structural DNA — short lines, white space, scannable on mobile. Never narrative essays. Never paragraphs of prose. Each number, each beat, each statement gets its own line.

Writing sequence:

  1. Write the hook exactly as locked — must create behavioral recognition ("that's me")
  2. Write the first line — sets up proof, under 10 words, opens what comes next. No elaboration on the hook. No transition language.
  3. Follow the arc section by section — match the golden example's delivery pattern, not a thesis-first lecture
  4. Each block: 1–3 sentences doing one thing
  5. Place line breaks for rhythm (see Line Break Decision Tree)
  6. Write the closer
  7. Write the conversation prompt — this is not an afterthought, it's the distribution engine
  8. Check total length: 1,300–1,900 characters target
  9. Read aloud — check for metronomic rhythm and AI patterns
  10. Check characters: one scenario per post, pronouns established before use, numbers consistent

Step 3: Self-Check Before Delivering

Structural comparison (check FIRST — before anything else):

Authority checks:

Distribution checks:

Status checks:

Step 4: Flag Known Weaknesses

After delivering, explicitly note what might need work:

Step 5: Revision + QC

Pass 1 — Copy QC (copy-qc.md): Scan every line against all 11 patterns. Fix every P1 and P2.

QC posture: fail-first. When a pattern is detected, the default is FAIL. To keep a flagged line, cite a specific golden example that uses the same structure and was annotated as acceptable in its Copy QC Notes. "I think it works here" is not sufficient — show the precedent or rewrite. If no golden example does the same thing, it fails.

Pass 2 — Sentence Editor (linkedin-sentence-editor.md): Audit every line against all 8 rules. Present changelog.

Only deliver as "ready to ship" after both passes are clean.

Line Break Decision Tree

Give it its own line when:

Keep it in a block when:

The staccato test: More than 3 consecutive single-sentence lines? Add a multi-sentence block.

Angle-Specific Build Notes

Reframe

Contrarian Take

Proof Story

Math Post

Identity Shift

Pattern Call-Out

First Principles

Lesson Learned

Common First-Pass Mistakes to Avoid

MistakeWhat to Do Instead
No conversation prompt — post ends on the closerThe closer is the takeaway. The conversation prompt is the distribution engine. Both are required.
Conversation prompt is "What do you think?" or "Agree?"Use one of the six prompt types. Specific, status-positive, requires multi-sentence answer.
Post is under 1,300 charactersAdd a beat. The algorithm treats short posts as low-effort. Even reframes need setup and implication.
Post is over 2,200 charactersCompress or split. Completion rate drops past this point. If it needs more room, it's a carousel or an article.
Hook announces a topic instead of creating a gapLead with a scene, claim, or number. Not "Let's talk about client retention."
Hook gives a conclusion instead of recognition"You don't know what your clients value" puts them on defense. "You already know which deliverables you'd cut first" creates recognition.
First line elaborates on the hookThe hook already landed. The first line opens proof (a list, steps, evidence). Under 10 words. No transition.
Post reads as narrative essay or dense paragraphsSteve's DNA: short lines, white space, scannable. Each number and beat gets its own line. Never paragraphs of prose.
Multiple characters or scenarios in one postOne scenario. One character. Escalate within that thread. If you need multiple examples, the article handles it.
Pronouns before establishing who they refer to"She was confident" — who? Establish role + action, then pronouns.
Post tries to make three pointsOne insight. One through-line. One conversation prompt.
Every sentence gets its own lineMix single-line beats with 2–3 sentence blocks. Staccato-everything is an AI pattern.
Post sounds like a motivational posterGround every insight in operational reality. voice.md vocabulary.
Post has an external link60% reach penalty. If you need to link, put it in a comment (though this is also being penalized in 2026).
Status angle not designedIf you can't articulate what engaging signals about the reader, the post may inform but won't earn engagement.
Closer tries to be an aphorismConcrete beats clever. "What they could see did." not "Perception is the ultimate currency."
Conversation prompt asks people to admit failureNobody wants their network to see them admitting weakness. Ask for expertise, not vulnerability.
Post is thesis-led instead of scene-ledThe golden examples all lead with scenes and let the insight emerge through narrative. If the hook announces the insight and the body illustrates it, the structure is backwards. Rewrite scene-first.
QC rationalizes flagged patterns instead of failing themWhen copy-qc detects a pattern, the default is FAIL. Cite a golden example precedent to keep it, or rewrite. No rationalizing.