linkedin-engagement

LinkedIn Engagement Skill

Configuration

LinkedIn profile: Lara Knuth (real profile) Company: fabrikIQ / Dresden AI Insights Focus: MES, OEE, manufacturing data analysis, SME digitalization

Target regions:

Primary: DACH (DE/AT/CH), USA, Canada
Secondary: UK, Nordics, Benelux
Tertiary: Japan, South Korea, Southeast Asia

Focus hashtags:

DE: #Fertigung #Industrie40 #OEE #MES #Digitalisierung #KMU #Qualitaetssicherung
EN: #Manufacturing #Industry40 #SmartFactory #MES #OEE #DigitalTransformation #LeanManufacturing

Slash Commands

/linkedin-post [region]

Purpose: Generates region-specific LinkedIn post

Parameters:

region: us | eu | asia (default: eu)
type: text | article-teaser | poll (default: text)
image: true | false (default: false)

Workflow:

Capture topic/key message
Select template based on region
Generate post with anti-AI detection
Optional: Generate image via Gemini
Add optimized hashtags
Show preview for approval

Output format:

## LinkedIn Post [Region: EU]

### Post-Text:
[Generierter Text]

### Hashtags (5):
#Hashtag1 #Hashtag2 ...

### Bild-Prompt (falls angefordert):
[Gemini-Prompt für Bildgenerierung]

### Beste Posting-Zeit:
[Region-spezifische Empfehlung]

### Checkliste:
- [ ] Kein AI-Slop?
- [ ] Erste 2 Zeilen = Hook?
- [ ] CTA vorhanden?

/linkedin-article [region]

Purpose: Creates LinkedIn article WITH teaser post

Workflow:

Capture article topic and key points
Generate article structure (800-1500 words)
Create separate teaser post (max 300 characters before "...more")
Generate header image via Gemini
Optimize SEO (title, description)

Output format:

## LinkedIn Artikel: [Titel]

### Teaser-Post (für Feed):
[Hook-Text, max 300 Zeichen]

[Link zum Artikel]

#Hashtags

---

### Artikel-Inhalt:

**Titel:** [SEO-optimiert]

**Intro:** [Hook, 2-3 Sätze]

**Hauptteil:**
[Strukturierter Content mit Zwischenüberschriften]

**Fazit:** [Call-to-Action]

---

### Header-Bild Prompt:
[Gemini-Prompt für 1200x627 Header]

### SEO-Daten:
- Titel: [max 60 Zeichen]
- Beschreibung: [max 160 Zeichen]
- Keywords: [...]

/linkedin-comment [url]

Purpose: Generates value-first comment for someone else's post

Workflow:

Load post content (via URL or description)
Analyze author region (name, language, company)
Generate comment adapted to region
Check anti-AI detection

Rules:

Value first, then (optionally) own experience
No direct advertising
Authentic reaction to content
Asking a question encourages engagement

Output format:

## Kommentar für: [Post-Titel/Autor]

**Autor-Region:** [geschätzt: US/EU/Asia]
**Ton-Empfehlung:** [Direct/Sachlich/Respektvoll]

### Vorgeschlagener Kommentar:
[Text, 50-150 Wörter]

### Alternative (kürzer):
[Text, 20-50 Wörter]

/linkedin-scan

Purpose: Scans relevant hashtags/influencers for engagement opportunities

Workflow:

Search hashtags: #Manufacturing, #MES, #OEE, #Industrie40
Identify posts with high engagement potential
Prioritize by: relevance, author reach, recency
Show top 10 with comment recommendation

Output format:

## LinkedIn Scan: [Datum]

### Engagement-Opportunities (Top 10)

1. **[Autor]** - [Titel/Hook]
   Reichweite: [geschätzt] | Engagement: [Likes/Comments]
   Region: [US/EU/Asia]
   → Kommentar-Empfehlung: [Kurz-Idee]

2. ...

### Trending Topics diese Woche:
- [Topic 1]: [Warum relevant]
- [Topic 2]: ...

/linkedin-monitor

Purpose: Monitors own posts for new comments, suggests replies

Workflow:

Load list of own posted content (from tracking.md)
Check each post for new comments
Analyze comment content and author
Generate reply suggestions

Output format:

## LinkedIn Monitor: [Datum]

### Neue Kommentare (seit letztem Check)

**Post:** [Post-Titel/Hook]
**Gepostet:** [Datum]
**Aktuelle Stats:** ♥ [Likes] | 💬 [Comments] | 🔄 [Shares]

#### Neuer Kommentar von [Name] ([Position]):
> "[Kommentar-Text]"

**Autor-Analyse:**
- Region: [US/EU/Asia]
- Relevanz: [Potentieller Lead/Peer/Troll]
- Ton: [Positiv/Neutral/Kritisch]

**Antwort-Vorschlag:**
[Generierte Antwort, regional angepasst]

**Alternative (kürzer):**
[Kürzere Version]

---

### Antwort-Priorität:
1. 🔴 DRINGEND: [Kritische Fragen, potentielle Leads]
2. 🟡 WICHTIG: [Fachliche Diskussionen]
3. 🟢 OPTIONAL: [Einfache Zustimmungen]

/linkedin-image [prompt]

Purpose: Generates LinkedIn-optimized image via Gemini

Integration with gemini-image-gen skill:

# Verwendet GEMINI_API_KEY aus .env.local
from google import genai
from google.genai import types
from dotenv import load_dotenv
load_dotenv('.env.local')

client = genai.Client(api_key=os.environ.get("GEMINI_API_KEY"))

# Modelle (funktionieren in DE!):
# - gemini-2.5-flash-image: Schnell, gut für einfache Grafiken
# - gemini-3-pro-image-preview: Höhere Qualität, komplexere Szenen (EMPFOHLEN)
# - imagen-4.0-generate-001: Fotorealistische Bilder

response = client.models.generate_content(
    model='gemini-3-pro-image-preview',  # FUNKTIONIERT IN DEUTSCHLAND
    contents=[prompt],
    config=types.GenerateContentConfig(
        response_modalities=['IMAGE']
    )
)

LinkedIn image formats:

Type	Size	Usage
Post image	1200x1200	Square, feed-optimized
Article header	1200x627	1.91:1 ratio
Carousel slide	1080x1080	PDF upload

Optimized prompts for manufacturing:

"Clean, professional infographic showing [TOPIC].
Modern flat design, blue and white color scheme,
minimal text, manufacturing/industrial context.
LinkedIn business style, 1200x1200px"

Output format:

## LinkedIn Bild generiert

**Prompt verwendet:**
[Optimierter Prompt]

**Modell:** gemini-2.5-flash-image
**Format:** 1200x1200 (Post) / 1200x627 (Article)

**Datei:** [Pfad zur generierten Datei]

**Verwendung:**
- [ ] Als Post-Bild hochladen
- [ ] Als Artikel-Header
- [ ] Für Carousel (weitere Slides nötig?)

/linkedin-analytics

Purpose: Shows performance overview of posted content

Metrics:

Impressions
Engagement Rate (Likes + Comments + Shares / Impressions)
Click-Through Rate (for articles)
Follower growth

Output format:

## LinkedIn Analytics: [Zeitraum]

### Top Performer

| Post | Datum | 👁 Impressions | ♥ Likes | 💬 Comments | ER% |
|------|-------|---------------|---------|-------------|-----|
| [Titel] | [Datum] | [X] | [Y] | [Z] | [%] |

### Insights:
- Beste Posting-Zeit: [Tag/Uhrzeit]
- Beste Content-Art: [Text/Artikel/Poll]
- Beste Hashtags: [Top 3]

### Empfehlungen:
- [Konkrete Handlungsempfehlung basierend auf Daten]

Regional templates

US/Canada template

Style: Direct, story-driven, personal brand Language: English Hashtags: 3-5, at the end

Structure:

[Hook - kontrovers oder überraschend, 1 Zeile]

[Leerzeile - wichtig für Mobile!]

[Personal Story mit konkreten Zahlen, 2-3 Sätze]

[Insight/Lesson, Bullet Points OK aber nicht genau 3]

[Vulnerable Admission - was ging schief]

[Soft CTA - Frage an Community]

#Manufacturing #MES #OEE #DigitalTransformation

Forbidden:

"I'm thrilled to announce"
"Excited to share"
"I'm humbled"
More than 5 emojis

Works:

Concrete numbers: "Reduced downtime by 23%"
Hot takes: "Unpopular opinion: MES is overkill for most SMBs"
Lessons learned with vulnerability
"Here's what I learned after..."

EU/DACH template

Style: Objective, fact-based, understatement Language: German or English (depending on target audience) Hashtags: 3-5, DE variants

Structure:

[Sachliche Eröffnung - Thema klar benennen]

[Kontext mit Daten/Zahlen aus echten Projekten]

[Pragmatischer Insight - was funktioniert, was nicht]

[Optional: Normen-Referenz (DIN, ISO, VDI)]

[Offene Frage - keine rhetorische]

#Fertigung #OEE #Industrie40 #MES #Digitalisierung

Forbidden:

Exaggerations ("revolutionary", "game-changer")
Too much self-promotion
American hustle culture
"I did everything right"

Works:

Case studies with methodology
Normative references
"This is how we did it" (team focus)
Critical reflection

Asia template (Japan, Korea, Southeast Asia)

Style: Respectful, relationship-first, indirect Language: English (international) Hashtags: 3-4, conservative

Structure:

[Höfliche Einleitung - Beobachtung, nicht Belehrung]

[Gemeinsames Lernen - "I noticed...", "I was impressed by..."]

[Eigene Erfahrung als Angebot, nicht als Expertise]

[Respektvolle Frage - Interesse an lokaler Perspektive]

#Manufacturing #Industry40 #QualityManagement

Forbidden:

Direct criticism
"I know better"
Starting business too quickly
Cultural stereotypes

Works:

Kaizen/Monozukuri as a connecting point (Japan)
Respect for local expertise
Long-term relationship building
Questions instead of statements

Human authenticity vs AI language

Scientifically documented AI markers (2024/2025 research)

Typical AI signals that must be AVOIDED:

Feature	Typical of AI	Human
Sentence length	Uniformly 15-20 words	Varies greatly: 3 to 40+ words
Paragraphs	Identical length	Chaotic, varied
Punctuation	Perfect, many em dashes (—)	Occasionally wrong, comma errors
Vocabulary	"Furthermore", "Moreover", "Delve"	Colloquial language, filler words
Structure	Always hook-body-CTA	Sometimes no conclusion, abrupt
Emotion	"I'm excited", "I'm thrilled"	Level-headed or genuinely frustrated
Lists	Exactly 3 or 5 points	2, 4, 7 - odd numbers OK
Formatting	Perfect markdown structure	Inconsistent

BUILD IN HUMAN IMPERFECTION

Deliberately set human markers (subtle, not exaggerated):

Spelling/typos (max 1-2 per post, natural):
- "Mitarbeier" instead of "Mitarbeiter"
- "das" vs "dass" mix-up
- Double letters: "Maschiene"
- Missing letter: "Frtigung"
- IMPORTANT: Only plausible typos, no deliberately dumb mistakes
Punctuation variants:
- Missing comma: "Aber naja das ist ein anderes Thema"
- Period instead of question mark: "Was denkt ihr."
- No period at the end (informal)
- Double space (happens while typing)
Grammar deviations (natural, not wrong):
- Sentence breaks: "Das Problem war... egal."
- Appended thoughts: "Achso, hab ich vergessen:"
- Colloquial language: "Ist halt so" instead of "Es ist so"
- Jumps in thought without transitions
Structural imperfection:
- Paragraphs of varying length (2 lines, then 5, then 1)
- Sometimes only 1 line as a paragraph
- No obvious pattern
- Abrupt ending OK (no forced CTA)
- Sometimes forgotten paragraph break

FORBIDDEN FORMATTING

Do NOT use ANY of these elements in posts:

Emojis as bullet points (rocket before point 1, etc.)
Perfect emoji lists with even spacing
Checkboxes as list items
Icons/symbols at the start of a line
Horizontal separators (--- or ===)
Headings in posts (## Title)
Code blocks or backticks
Perfectly aligned tables
Bold for every important term

ALLOWED (very sparingly):

1-2 emojis at the end or as an accent (not in every post)
Regular numbers for lists (1. 2. 3.)
Bullet points without emojis (- point)
Occasionally an arrow (->)

AI PHRASES: TOTAL BLACKLIST

German AI markers:

"In der heutigen Zeit"
"Wie wir alle wissen"
"Es ist allgemein bekannt"
"Zusammenfassend lässt sich sagen"
"Es bleibt festzuhalten"
"Abschließend möchte ich betonen"
"Dies führt uns zu der Erkenntnis"
"In diesem Zusammenhang"
"Darüber hinaus"
"Des Weiteren"
"Schlussendlich"
"Es ist von entscheidender Bedeutung"

English AI markers:

"Delve into" / "Delve deeper"
"Leverage synergies"
"In today's fast-paced world"
"It's important to note that"
"Furthermore" / "Moreover" / "Additionally"
"This begs the question"
"Needless to say"
"At the end of the day"
"Game-changer" / "Revolutionary"
"Seamlessly integrate"
"Navigate the complexities"
"Unlock the potential"
"Fostering innovation"

Em dash overuse (—):

AI uses em dashes excessively between parts of sentences
People tend to use hyphens (-) or simply commas
Or parentheses (like here)
Max 1 em dash per post, if any

AUTHENTIC ALTERNATIVES

AI phrase	Human alternative
"I'm thrilled to announce"	"Endlich fertig:" or straight into the topic
"Here are 5 key takeaways"	"Was hat funktioniert?"
"Let me share my journey"	"Kurze Geschichte dazu:"
"This changed everything"	"Hat geholfen"
"You won't believe"	"Überraschend war:"
"In der heutigen Zeit"	"Momentan" / "Gerade" / omit
"Zusammenfassend"	"Also:" / "Heißt:"
"Darüber hinaus"	"Außerdem" / "Und noch:"
"Key learnings"	"Was wir gelernt haben"
"Moving forward"	omit or "next up"
"It's worth noting"	just say it directly
"I'm excited to"	omit, get straight to the point

EXAMPLES: AI vs HUMAN

AI-generated (BAD):

I'm thrilled to announce that we've successfully implemented a new OEE tracking system!

Here are 5 key takeaways:

Punkt 1 Real-time visibility
Punkt 2 23% efficiency gain
Punkt 3 Data-driven decisions
Punkt 4 Seamless integration
Punkt 5 Team empowerment

This truly changed everything for us. What's your experience with OEE?

#Manufacturing #Industry40 #OEE #DigitalTransformation #SmartFactory

Human-written (GOOD):

Letzte Woche endlich unser OEE-Tracking live geschaltet.

Ehrlich gesagt war der Weg dahin holprig - erste Excel-Version hat keiner benutzt, zu kompliziert. Jetzt simpler und die Schichtleiter tragen tatsächlich ein.

23% weniger Stillstand seit dem. Ob das so bleibt weiß ich nicht, Anfangs-Effekt vielleicht

Was mich interessiert: wie kriegt ihr eure Leute dazu, sowas auch wirklich zu nutzen.

#Fertigung #OEE #MES

Anti-AI-Detection Rules (LinkedIn-specific)

FORBIDDEN (LinkedIn AI slop):

Phrases:

"I'm thrilled to announce..."
"Excited to share that..."
"I'm humbled and honored..."
"Here are 5 key takeaways:"
"Let me tell you a story..."
"If you found this valuable, like and share"
"Agree? 👇"
"DM me 'STRATEGY' for..."
"I helped 100+ companies achieve..."
"This changed everything for me"
"You won't believe what happened next"

Structures:

Exactly 5 bullet points
Perfect emoji lines (🎯 Point 1 / 🚀 Point 2)
Identical paragraph lengths
"Hook → Story → Lesson → CTA" too obvious
Every sentence a new line (poetry-style spam)

Emojis:

🚀🔥💡🎯💪 combo
More than 3-4 per post
Emoji at the start of a line (list)

AUTHENTICITY SIGNALS:

Linguistic:

Varying sentence lengths - short. Then longer, because the thought needs it.
Incomplete thoughts - "But oh well, that's another topic."
Regional expressions - DE: "Naja", "halt", "irgendwie" / US: "tbh", "ngl"
Subsequent corrections - "Edit: Forgot to mention..."
Genuine questions without an obvious answer

Content-related:

Specific contexts instead of generic claims
Admit mistakes - "Our first attempt was a flop"
Nuanced opinions - "It depends..."
Local references (trade fairs, associations, cities)
Temporal context - "Last week with a client in Saxony..."

Structural:

Not every post needs a CTA
Sometimes just a question, no answer
Paragraphs of different lengths
Occasional typos (max 1-2)

Hashtag strategy

German (DACH)

Reach	Hashtags
High (>100k)	#Industrie40 #Digitalisierung #KMU
Medium (10-100k)	#Fertigung #OEE #MES #Produktion
Niche (<10k)	#Qualitaetssicherung #Maschinendaten #SmartFactory

Recommendation: 1 high + 2 medium + 2 niche = 5 hashtags

English (International)

Reach	Hashtags
High (>500k)	#Manufacturing #Industry40 #DigitalTransformation
Medium (50-500k)	#SmartFactory #LeanManufacturing #OEE
Niche (<50k)	#MES #ManufacturingExcellence #ShopFloor

Hashtag Rules:

Hashtags at the end of the post (not inline)
No hashtags in the first paragraph (disrupts the hook)
Max 5 hashtags (more = spammy)
Mix of reach tiers
No invented hashtags

Posting Times

Optimal by region:

Region	Best Days	Best Times (local)
DACH	Tue-Thu	08:00-09:00, 17:00-18:00
USA East	Tue-Thu	08:00-10:00, 17:00-18:00
USA West	Tue-Thu	07:00-09:00, 16:00-17:00
UK	Tue-Thu	08:00-09:00, 17:00-18:00
Asia	Wed-Fri	09:00-11:00 (local time)

Avoid:

Monday morning (too much noise)
Friday afternoon (weekend mode)
Weekend (except Sunday evening for Monday visibility)

Article Teaser Formula

Hook structure (max 300 characters before "...more"):

[Provokante These oder überraschende Zahl]

[1 Satz Kontext]

[Neugier wecken: "Im Artikel zeige ich..." oder "3 Dinge, die wir gelernt haben:"]

Example:

85% der OEE-Implementierungen liefern nicht den erwarteten ROI.

Wir haben 12 Projekte analysiert und die 3 häufigsten Fehler identifiziert.

Im Artikel: Konkrete Zahlen und wie ihr sie vermeidet 👇

[LINK]

#OEE #Manufacturing #Fertigung

Tracking Log

Published Content

Date	Type	Title/Hook	Region	URL	Status
[Date]	Post/Article	[Short title]	EU/US	[URL]	✅ Posted

Performance History

Date	Post	👁 Impressions	♥ Likes	💬 Comments	🔄 Shares	ER%
[Date]	[Title]	[X]	[Y]	[Z]	[W]	[%]

Comment Queue (own posts)

Post	New comment from	Content (short)	Answered?
[Title]	[Name]	[Short quote]	⏳/✅

Gemini Integration for Images

Setup

# Environment Variable setzen
export GOOGLE_AI_API_KEY="your-key-here"

# Dependencies
pip install google-genai pillow python-dotenv

LinkedIn-optimized Prompts

Infographic (OEE/Data):

Professional infographic showing OEE calculation breakdown.
Clean flat design, blue (#0077B5 LinkedIn blue) and white.
Icons for Availability, Performance, Quality.
Minimal text, data visualization style.
1200x1200px, white background.

Header for Article:

Modern manufacturing facility abstract visualization.
Digital data overlay, blue tones, professional look.
No text, suitable for article header.
1200x627px, LinkedIn article format.

Carousel Slide:

Single slide for LinkedIn carousel about [TOPIC].
Large bold headline area, clean infographic style.
Blue and white, professional B2B manufacturing.
1080x1080px square format.

Generation Workflow:

/linkedin-image [Thema]
Skill generates optimized prompt
Call to Gemini API
Save with timestamp
Display preview + usage notes

Persona: Lara Knuth

LinkedIn Profile:

Name: Lara Knuth (real)
Position: Founder fabrikIQ / Dresden AI Insights
Location: Dresden, Saxony
Background: MES expert, COO/CEO experience in SMEs

Authenticity Elements:

Twins (lack of time, multitasking - mention sparingly)
Practical experience from real projects
Saxony/DACH perspective
Critical of hype, pragmatic

Writing style LinkedIn:

More professional than Reddit, but not stiff
German for DACH, English for international
Numbers and facts, but with storytelling
Admit mistakes, share learnings

Technical implementation (web access)

Experiences from Reddit skill (transferable)

Problem: WebFetch is blocked by many platforms Solution: Browser automation or curl with user agent

LinkedIn-specific challenges

Aspect	Reddit	LinkedIn
Public API	✅ `/new.json`	❌ None
Scraping allowed	⚠️ Rate-limited	❌ Strictly prohibited (ToS)
Auth required	Only for posts	For almost everything
Bot detection	Moderate	Aggressive

Recommended methods (priority)

1. Claude-in-Chrome MCP (BEST OPTION)

Advantages:

Uses real browser session (logged in)
No scraping suspicion
Full access to feed, comments, analytics

Step-by-step workflows:

/linkedin-scan workflow

1. tabs_context_mcp (createIfEmpty: true)
   → Prüft ob Tab-Gruppe existiert

2. tabs_create_mcp
   → Neuen Tab erstellen

3. navigate (url: "https://www.linkedin.com/feed/hashtag/manufacturing")
   → Hashtag-Feed öffnen

4. browser_wait_for (time: 3)
   → Warten bis Feed geladen

5. read_page (tabId: X, filter: "all")
   → Accessibility-Tree extrahieren

6. find (query: "post with reactions", tabId: X)
   → Posts mit Engagement finden

7. Analyse und Kommentar-Empfehlungen generieren

/linkedin-comment workflow

1. navigate (url: "[POST-URL]")
   → Direkt zum Post navigieren

2. read_page (tabId: X)
   → Post-Inhalt und Autor extrahieren

3. get_page_text (tabId: X)
   → Volltext für Kontext

4. Kommentar generieren (siehe Templates)

5. find (query: "comment input field", tabId: X)
   → Kommentarfeld finden (für manuelle Eingabe)

/linkedin-monitor workflow

1. navigate (url: "https://www.linkedin.com/in/lara-knuth/recent-activity/")
   → Eigene Aktivitäten öffnen

2. read_page → Posts mit Kommentar-Counts finden

3. Für jeden Post mit neuen Kommentaren:
   - navigate → Post öffnen
   - read_page → Kommentare extrahieren
   - Antwort-Vorschläge generieren

/linkedin-analytics workflow

1. navigate (url: "https://www.linkedin.com/analytics/")
   → Analytics-Dashboard öffnen

2. computer (action: "screenshot", tabId: X)
   → Screenshot für visuelle Analyse

3. read_page → Zahlen extrahieren

4. Performance-Report generieren

2. Manual copy-paste (FALLBACK)

If Claude-in-Chrome is not available:

User: Hier ist der Post-Inhalt: [PASTE]
Claude: Analysiert und generiert Kommentar

3. curl with user agent (LIMITED)

Works ONLY for public articles (not feed):

# Öffentlicher LinkedIn-Artikel (ohne Login sichtbar)
curl -s -A "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36" \
  "https://www.linkedin.com/pulse/[article-slug]" | head -1000

Limitation: Feed posts, comments, analytics = NOT accessible without login

4. LinkedIn API (if OAuth configured)

# Prüfen ob LinkedIn-Token vorhanden
echo $LINKEDIN_ACCESS_TOKEN

# API-Aufruf (nur mit gültiger OAuth-App)
curl -H "Authorization: Bearer $LINKEDIN_ACCESS_TOKEN" \
  "https://api.linkedin.com/v2/me"

Status: Currently NOT configured. LinkedIn API requires:

LinkedIn Developer App
OAuth 2.0 flow
Approval for Marketing API (for posts)

Implementation per command

Command	Best method	Fallback
`/linkedin-scan`	Claude-in-Chrome	Manual (user shows feed)
`/linkedin-comment [url]`	Claude-in-Chrome → open URL	User pastes post text
`/linkedin-monitor`	Claude-in-Chrome → own posts	User reports comments
`/linkedin-analytics`	Claude-in-Chrome → analytics tab	User shares screenshot

Code snippets for Claude-in-Chrome

Extract post content:

// Via mcp__claude-in-chrome__javascript_tool
const posts = document.querySelectorAll('[data-urn*="activity"]');
const postData = Array.from(posts).map(p => ({
  author: p.querySelector('.update-components-actor__name')?.innerText,
  text: p.querySelector('.feed-shared-update-v2__description')?.innerText,
  reactions: p.querySelector('.social-details-social-counts__reactions-count')?.innerText
}));
return JSON.stringify(postData, null, 2);

Load hashtag feed:

// mcp__claude-in-chrome__navigate
url: "https://www.linkedin.com/feed/hashtag/manufacturing"

Important: No automation of posts/comments!

The skill only generates suggestions. The actual posting/commenting:

Must be done manually by user
LinkedIn ToS prohibit bot posts
Account suspension for automation

Quality checklist before posting

Post/Article:

No AI slop phrases?
Hook in first 2 lines?
Vary sentence lengths?
Authentic voice (Lara)?
Regionally appropriate (US/EU/Asia)?
Hashtags at the end (max 5)?
Image if it makes sense?
No exaggerated claims?

Comment:

Value-first (not advertising)?
Appropriate for the author's region?
Under 150 words?
Genuine reaction to content?

Reply to own posts:

Promptly (< 24h)?
Personal, not generic?
Furthering the discussion?
When facing criticism: stay objective?

About this skill

Configuration

Slash Commands

/linkedin-post [region]

/linkedin-article [region]

/linkedin-comment [url]

/linkedin-scan

/linkedin-monitor

/linkedin-image [prompt]

/linkedin-analytics

Regional templates

US/Canada template

EU/DACH template

Asia template (Japan, Korea, Southeast Asia)

Human authenticity vs AI language

Scientifically documented AI markers (2024/2025 research)

BUILD IN HUMAN IMPERFECTION

FORBIDDEN FORMATTING

AI PHRASES: TOTAL BLACKLIST

AUTHENTIC ALTERNATIVES

EXAMPLES: AI vs HUMAN

Anti-AI-Detection Rules (LinkedIn-specific)

FORBIDDEN (LinkedIn AI slop):

AUTHENTICITY SIGNALS:

Hashtag strategy

German (DACH)

English (International)

Hashtag Rules:

Posting Times

Optimal by region:

Avoid:

Article Teaser Formula

Hook structure (max 300 characters before "...more"):

Example:

Tracking Log

Published Content

Performance History

Comment Queue (own posts)

Gemini Integration for Images

Setup

LinkedIn-optimized Prompts

Generation Workflow:

Persona: Lara Knuth

Technical implementation (web access)

Experiences from Reddit skill (transferable)

LinkedIn-specific challenges

Recommended methods (priority)

1. Claude-in-Chrome MCP (BEST OPTION)

/linkedin-scan workflow

/linkedin-comment workflow

/linkedin-monitor workflow

/linkedin-analytics workflow

2. Manual copy-paste (FALLBACK)

3. curl with user agent (LIMITED)

4. LinkedIn API (if OAuth configured)

Implementation per command

Code snippets for Claude-in-Chrome

Important: No automation of posts/comments!

Quality checklist before posting

Post/Article:

Comment:

Reply to own posts: