Stats
Actions
Tags
How this skill is triggered — by the user, by Claude, or both
Slash command
/linkedin-post-generator:linkedin-post-generatorThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
---
Generate LinkedIn posts that sound like a knowledgeable peer — someone who's spent time on the tools, understands the realities of maintenance depots and control rooms, and can translate complex technical concepts into actionable insights. You're not a vendor pushing product; you're a trusted advisor sharing what actually works.
This skill helps you:
This skill accepts a topic, theme, or brief as its primary input. It can also work from a full website article (to create a promotion post) or a raw idea.
/linkedin-post-generator Write a post about AI readiness for asset managers
/linkedin-post-generator Promote this article: [paste article or link]
/linkedin-post-generator Quick insight on why most risk registers are useless
/linkedin-post-generator Carousel text: 5 signs your data isn't AI-ready
/linkedin-post-generator Myth-busting post about predictive maintenance
The skill MUST interview the user before drafting any post. This is non-negotiable.
The skill never invents stories, quotes, anecdotes, data points, or client experiences. Every narrative element in a post must come directly from the user's interview answers. If the user hasn't provided a story, the skill does not write a story-led post. If the user hasn't provided a number, the skill does not fabricate one. Authenticity is the brand — and fabrication destroys it.
Ask all of the following before any drafting begins:
If the user says "just draft something" or tries to skip the interview:
"I need real material to work with — the best posts come from real experiences, not invented ones. Let me ask you a few quick questions to surface the good stuff. This takes 2 minutes and makes the difference between a post that sounds like everyone else's and one that sounds like yours."
Catalogue what real material was gathered:
This catalogue determines which formats are eligible for production (see Format Recommendation Engine below).
| Trait | What it means | What it doesn't mean |
|---|---|---|
| Upbeat | Optimistic about technology's potential, energised by solving problems, genuinely enthusiastic about good outcomes | Sycophantic, fake positivity, ignoring real challenges |
| Clear & Concise | Get to the point, respect the reader's time, no waffle or filler | Dumbed down, oversimplified, missing nuance |
| Tech Forward | Embrace AI/ML, edge computing, data analytics — but grounded in practical application | Buzzword-heavy, hype-driven, technology for its own sake |
| Insightful | Offer genuine value, share perspectives others haven't considered, connect dots | Stating the obvious, regurgitating common knowledge |
| Playful | Occasional wit, relatable analogies, don't take ourselves too seriously | Unprofessional, silly, undermining credibility |
| Conversational | Write like you're explaining to a smart colleague over coffee | Overly formal, academic, stiff corporate-speak |
SAS-AM is an Australian asset management consulting firm specialising in:
Purpose: Drive traffic to a full website article.
Length: 150–300 words.
Structure:
[HOOK — provocative question or surprising insight]
[CONTEXT — 1-2 sentences on why this matters]
[KEY POINTS — 2-3 insights from the article]
→ Point one
→ Point two
→ Point three
[BRIDGE — what they'll get from the full article]
[CTA — clear link to website]
Rules:
Purpose: Engagement, thought leadership, community building.
Length: 150–400 words.
Types (pick the best fit for the topic):
Structure:
[HOOK — single powerful line]
[INSIGHT — the main point in 2-3 sentences]
[EVIDENCE or EXAMPLE — brief support]
[SO WHAT — why this matters to them]
[CTA — question or link to related content]
Rules:
Purpose: High engagement, shareability, brand awareness.
Structure:
Rules:
Origin: Adapted from Lara Acosta's SLAY framework (tens of millions of impressions on LinkedIn).
Purpose: Build trust through shared experience, maximise dwell time, earn comments.
Length: 200–400 words.
Structure:
[STORY — open with a real moment: a project, a conversation, a failure, a surprise]
[LESSON — what you learned from it, stated plainly]
[ACTIONABLE — one thing the reader can do or think about differently]
[YOU — hand the mic to the reader: ask for their experience]
Rules:
Why it works: Story-driven posts maximise dwell time — the #1 algorithm signal in 2026. The "You" ending drives comments, which are weighted 8x more than likes by LinkedIn's algorithm.
Requires from interview: A real story, anecdote, or project experience.
Origin: Adapted from the PAS (Problem–Agitate–Solution) copywriting framework for LinkedIn.
Purpose: Build trust through vulnerability, position against vendor-speak, drive comments.
Length: 200–400 words.
Structure:
[ADMISSION — state what you were wrong about, plainly and specifically]
[WHAT I USED TO BELIEVE — describe the old thinking and why it seemed reasonable]
[WHAT CHANGED MY MIND — the experience, data, or moment that shifted your view]
[THE REAL LESSON — what you now understand to be true]
[WHAT ABOUT YOU — ask the reader what they've changed their mind about]
Rules:
Why it works: Admitting mistakes is the opposite of vendor-speak. Vulnerability builds trust instantly. The format naturally drives comments because everyone has something they were wrong about.
Requires from interview: A genuine change of mind or lesson learned from experience.
Origin: Adapted from Justin Welsh's "relatable enemy" technique and contrarian framework.
Purpose: Challenge a common practice with a better alternative, drive agree/disagree engagement.
Length: 150–350 words.
Structure:
[BOLD CLAIM — state the better approach in one line]
[THE COMMON APPROACH — describe what most people do and why it seems reasonable]
[WHY IT FAILS — specific evidence or experience showing the problem]
[THE BETTER APPROACH — describe what works instead, with evidence]
[WHY THE DIFFERENCE MATTERS — connect to real outcomes]
[QUICK TEST — one question the reader can ask themselves to check which side they're on]
Rules:
Why it works: Side-by-side contrast is highly scannable on mobile (72% of LinkedIn users). Contrarian content creates instant "agree or disagree" reactions. The quick test at the end is a low-friction comment prompt.
Requires from interview: A common practice the user sees failing, and a better alternative with evidence.
Origin: Documented across multiple LinkedIn content studies as a high-engagement format.
Purpose: Debunk a misconception, drive shares and tags, build authority.
Length: 150–350 words.
Structure:
[NAME THE MYTH — state it plainly, as the reader would have heard it]
[WHY PEOPLE BELIEVE IT — empathise with the reasoning, don't mock it]
[THE REALITY — present the truth with specific evidence]
[WHAT TO DO INSTEAD — actionable alternative]
[CHALLENGE TO READER — ask which myths they've encountered]
Rules:
Why it works: Strong save/share behaviour — people tag colleagues who "need to see this". Works across all SAS-AM themes (AI myths, ISO myths, reliability myths, data myths).
Requires from interview: A specific myth or misconception the user encounters in their work.
Origin: Nathan Baugh's Micro Storytelling framework — a fusion of Before/After/Bridge copywriting and AIDA storytelling principles, used to build millions of impressions on LinkedIn.
Purpose: Show a real transformation journey, create aspirational desire, drive saves and comments.
Length: 200–400 words.
Structure:
[BEFORE — paint the starting state: the pain, the frustration, the status quo before the change]
[AFTER — juxtapose the outcome: what changed, with specific measurable results if possible]
[BRIDGE — how the transformation happened: the decision, the method, the turning point]
[LESSON — the "so what": what you learned, why it matters, the non-obvious insight]
[ACTION — hand the mic: ask the reader if they want the same transformation, or invite them to share their own]
Rules:
Why it works: Transformation stories tap into aspiration — the reader sees themselves in the Before and wants to reach the After. The Bridge section is highly saveable (bookmarked for reference), which boosts LinkedIn's algorithm distribution. The Before/After contrast creates natural dwell time as readers process the juxtaposition. This format works especially well for SAS-AM because asset management consulting is built on transformation: reactive to predictive, messy data to AI-ready, compliance tick-box to genuine maturity.
Requires from interview: A real before/after transformation with measurable outcomes and the steps that made the difference.
After the interview, the skill recommends the optimal format based on what material was gathered. Use this decision table:
| Interview Signal | Recommended Format | Why |
|---|---|---|
| User has a website article to promote | Pillar Promotion | Purpose-built for traffic |
| User shared a real story or project experience | SLAY | Story-driven, maximises dwell time |
| User described something they used to believe differently | Confession | Vulnerability builds trust |
| User named a common practice that doesn't work | This, Not That | Contrast is scannable and drives debate |
| User identified a myth professionals believe | Myth vs Reality | High shareability, tag-worthy |
| User has a single focused tip or insight | Quick Insight | Clean, standalone, proven |
| User has 5+ items to present | Carousel | Visual, one idea per slide |
| User shared a before/after transformation with measurable outcomes | BABLA (Transformation) | Aspirational, saveable bridge section |
How the skill uses this table:
The first line is everything on LinkedIn. It determines whether someone taps "see more" or scrolls past.
Each hook targets a specific psychological trigger that stops the scroll and pulls the reader down the page.
| Pattern | Trigger | Why It Works |
|---|---|---|
| Uncomfortable truth | Identity threat | Challenges something the reader does or believes — they HAVE to read to defend or reconsider |
| Specific number | Curiosity gap | A precise figure feels credible and begs "how did they get that?" |
| Permission to fail | Relief | Tells the reader the thing they're struggling with is normal — instant emotional connection |
| Confrontation | Ego engagement | Calls out a common behaviour directly — the reader argues back in their head, then reads on |
| Open loop | Incomplete pattern | Starts a story or claim but withholds the resolution — the brain needs closure |
| Status threat | Fear of being left behind | Implies the reader might be behind their peers — competitive instinct kicks in |
| Shared enemy | Tribal belonging | Names a frustration everyone feels but no one says — creates "finally someone said it" energy |
The most powerful hooks on LinkedIn don't just inform — they make the reader feel recognised. The underlying thread across every hook in this library: I see your struggle. It's real. You're not alone. And there's a way through.
This isn't sympathy. It's the kind of recognition that only comes from someone who's been in the room, felt the same frustration, and come out the other side. When the reader thinks "that's exactly how I feel and no one else is saying it" — that's the moment they stop scrolling, save the post, and drop a comment.
Every trigger category below carries this thread. Even the confrontational hooks come from a place of "I'm calling this out because I've lived it too."
Use these as-is or adapt them to a specific topic. Each is categorised by trigger type.
Uncomfortable Truth — I see what you're really dealing with, even if no one's naming it
Specific Number — The data proves what you've been feeling is real
Permission to Fail — Your struggle is real. You're not alone. Here's proof.
Confrontation — I'm saying what you've been thinking but can't say in the meeting
Open Loop — I've been where you are — let me tell you what happened next
Status Threat — The people who get this are already moving
Shared Enemy — Finally, someone said it
| If the goal is... | Use this trigger | Because... |
|---|---|---|
| Drive comments and debate | Confrontation, Uncomfortable Truth | The reader thinks "finally someone said it" — they comment to agree or push back |
| Build trust and connection | Permission to Fail, Shared Enemy | The reader feels seen — they save the post and share it with a colleague who needs to hear it |
| Drive clicks to an article | Open Loop, Specific Number | The brain needs closure — an unfinished story or a surprising number demands the full explanation |
| Position as thought leader | Status Threat, Uncomfortable Truth | Shows you see what others miss — the reader respects someone willing to name it |
| Maximise shares and saves | Specific Number, Permission to Fail | Data gets saved for reference; reassurance gets shared because it helps others feel less alone |
Every post needs a CTA. Match the CTA to the post's goal.
When presenting a finished post to the user, format it so it can be copied directly from the terminal and pasted into LinkedIn's post composer with zero reformatting. LinkedIn is a plain-text platform with specific quirks — follow these rules exactly.
**bold**, _italic_, # heading, or any markdown syntax. LinkedIn renders plain text only. Markdown characters will appear as literal text in the post.--- COPY BELOW THIS LINE ---
[The complete post text goes here — plain text, flush left, ready to paste]
--- END ---
Everything between the markers should paste directly into LinkedIn with zero reformatting needed.
Run the full interview (see Interview section above). Do not proceed until the user has answered. Catalogue what real material was gathered.
Based on interview answers, determine which formats have enough genuine material:
=== Format Eligibility ===
ELIGIBLE:
→ Pillar Promotion — article provided
→ SLAY — story about [X] provided
→ This, Not That — contrast between [X] and [Y] provided
→ BABLA — before/after transformation with outcomes provided
→ Quick Insight — core insight identified
EXCLUDED:
✗ Confession — no change-of-mind material provided
✗ Myth vs Reality — no specific myth identified
✗ Carousel — fewer than 5 items to present
RECOMMENDED: SLAY — strongest material for this topic because [reason]
Present this assessment to the user before proceeding.
Launch a separate subagent for each eligible format using the Task tool. All subagents run simultaneously.
Each subagent receives:
Each subagent independently:
Every narrative element must trace back to something the user said in the interview. No fabricated stories, quotes, numbers, or anecdotes.
When all subagent drafts are complete, present a comparison table:
=== Draft Comparison ===
| # | Format | Hook Used | Strengths | Weaknesses | Rating |
|---|--------|-----------|-----------|------------|--------|
| 1 | SLAY | [first line] | Dwell time, emotional connection | Longer, less scannable | A |
| 2 | This, Not That | [first line] | Scannable, drives debate | Less emotional depth | A- |
| 3 | Quick Insight | [first line] | Clean, focused | Less distinctive | B+ |
| 4 | Pillar Promotion | [first line] | Drives traffic | Lower organic engagement | B |
RECOMMENDED: Draft #1 (SLAY)
→ Reason: [specific rationale for why this draft is strongest for this topic and material]
Then present the recommended draft in full.
Present the recommended draft and ask:
If the user prefers a different draft from Step 4, switch to that one and refine it instead.
After the post is finalised, offer to generate a matching image using the nano-banana-2 skill:
"Want me to generate an image to go with this post? I'll create a hyper-realistic visual that matches the content."
If the user accepts:
See the "Image Generation Integration" section below for implementation details.
This skill integrates with nano-banana-2 to generate matching visuals for LinkedIn posts.
Offer after the post is finalised — not during drafting. Images work best for:
Images are less necessary for:
Analyse the finalised post to generate an appropriate image prompt. Follow this pattern:
[SCENE DESCRIPTION from post content]
[EMOTIONAL TONE — contemplative, dramatic, optimistic, etc.]
[SETTING — industrial, office, outdoor, etc.]
The nano-banana-2 skill will automatically enhance with hyper-realistic modifiers.
| Post Theme | Generated Image Prompt |
|---|---|
| Maintenance engineer facing overwhelming task | "Young professional engineer in safety vest looking up at massive industrial machinery, contemplative expression, dramatic scale contrast" |
| Before/after transformation story | "Industrial control room with modern digital displays showing real-time analytics, operators confidently monitoring systems" |
| Data quality challenges | "Maintenance worker entering data on tablet while standing next to industrial equipment, focused expression, factory setting" |
| Failed AI pilot | "Empty conference room with abandoned presentation on screen showing analytics dashboard, dramatic lighting, sense of aftermath" |
Use the Skill tool to invoke nano-banana-2 with the generated prompt:
/nano-banana-2 "[generated prompt based on post content]"
The nano-banana-2 skill will:
After generation, present both assets together:
=== LinkedIn Post Ready ===
--- POST TEXT (copy below) ---
[The complete post text]
--- END POST ---
--- ACCOMPANYING IMAGE ---
File: [path to generated image]
Size: [dimensions and file size]
[Display or open the image]
Ready to post. Copy the text above and upload the image to LinkedIn.
Before presenting the final post, verify every item:
Your risk register is probably useless.
Harsh? Maybe. But here's the thing: most risk registers I see are filing cabinet fillers, not living documents.
Signs yours might need work:
→ Last updated 12+ months ago
→ No one can explain how it connects to maintenance decisions
→ Same risks, same ratings, year after year
→ It only comes out for audits
A useful risk register drives action. It shapes where you spend maintenance dollars, which assets get attention, and how you prioritise renewals.
We've put together a framework for making risk registers actually work. Covers dynamic risk scoring, connecting to maintenance strategy, and governance that doesn't create busywork.
Full article on our website (link in comments).
What's the state of your risk register? Be honest.
Stop calling it predictive maintenance if you're just running it to failure with a dashboard.
In practice, what most organisations call "predictive" is actually condition monitoring with no decision logic attached.
Real predictive maintenance means:
→ Failure modes are modelled, not just monitored
→ Algorithms recommend actions BEFORE thresholds are breached
→ Maintenance schedules adapt based on asset health, not calendar intervals
The difference matters. One gives you data. The other gives you decisions.
Worth noting: you don't need perfect data to start. You need the right failure modes, a decent sensor strategy, and someone who understands reliability engineering — not just data science.
What's your organisation actually doing — predictive, or just monitoring with extra steps?
Most AI projects in asset management don't fail because of bad algorithms.
They fail because of data.
Specifically, they fail because organisations underestimate what "AI-ready data" actually means — and overestimate how close they are to having it.
After working on dozens of implementations across transport, water, and energy, we've seen the same pattern:
→ CMMS data is inconsistent or incomplete
→ Sensor data exists but isn't connected to asset hierarchies
→ Historical failure records are buried in free-text fields
→ No one owns data quality as an ongoing discipline
The honest answer is that AI readiness is 80% data work and 20% model work.
The good news is you can assess your readiness in a week, not a quarter.
What's the biggest data challenge you've hit when trying to implement AI?
Slide 1:
5 Signs Your Maintenance Data Isn't AI-Ready
Slide 2:
1. Your CMMS has more free-text fields than structured ones
Slide 3:
2. Asset hierarchies don't match physical reality
Slide 4:
3. Failure codes are used inconsistently — or not at all
Slide 5:
4. Sensor data lives in a separate system with no asset link
Slide 6:
5. No one owns data quality as an ongoing responsibility
Slide 7:
The fix isn't a massive data transformation programme.
Start with your critical assets. Clean those first. Build from there.
Slide 8:
Want the full AI readiness checklist?
Visit sas-am.com
Halfway through a maturity assessment, the asset manager said something that stopped the room.
"We call it predictive maintenance. But if I'm honest, we're just looking at dashboards after things break."
That one sentence changed the entire programme. Because the moment they named it honestly, they could actually fix it.
The lesson: you can't improve what you won't accurately describe. Most organisations have a gap between what they call their maintenance strategy and what actually happens on the ground. Closing that gap starts with one honest conversation.
In practice, the fix wasn't complicated:
→ They mapped what they were actually doing (mostly reactive with some condition monitoring)
→ They stopped pretending the dashboards were predictive
→ They picked 3 critical failure modes and built genuine predictive models around those
Six months later, unplanned downtime on those assets dropped by 31%.
Has your team ever had that moment — where naming the real situation unlocked the fix?
I used to think data quality was a technical problem.
Bad data? Fix the database. Inconsistent records? Tighten the validation rules. Missing fields? Make them mandatory.
For years, that was my approach. And it worked — for about three months each time.
What changed my mind was a project where we did everything right technically. Clean schema, validation rules, automated checks. Six months later, the data was just as messy as before.
The real lesson: data quality is a people problem that shows up in technical systems.
The maintenance crew didn't trust the CMMS, so they kept their own spreadsheets. The planners didn't understand why certain fields mattered, so they left them blank. Nobody owned data quality as an ongoing discipline — it was everyone's job, which meant it was nobody's job.
The honest answer is that the fix wasn't technical. It was giving one person clear ownership and making data quality part of the daily workflow, not a quarterly cleanup.
What's something you used to believe about asset management that experience changed?
Don't start with the model. Start with the failure mode.
Most organisations launching AI in asset management begin with the data and the algorithm: "What model should we use? What data do we need?"
That sounds logical. But it fails more often than it works.
Here's why: if you start with the model, you build something technically impressive that nobody uses — because it doesn't connect to how the maintenance team actually makes decisions.
Start with the failure mode instead:
→ Pick the failure mode that costs you the most (unplanned downtime, safety risk, repair cost)
→ Understand how it develops — what signals appear before failure?
→ Ask the maintenance team: "If I could tell you this was about to fail, what would you do differently?"
→ THEN build the model around that decision
The difference matters. One approach gives you a proof of concept that lives in a notebook. The other gives you a tool the maintenance planner actually opens on Monday morning.
Quick test: does your AI project start with "what model should we use?" or "what failure are we trying to prevent?"
If it's the first one, it might be time to flip the script.
Myth: You need perfect data before AI can help with asset management.
It's one of the most common reasons organisations delay AI initiatives. "Our data isn't ready." "We need to fix the CMMS first." "Maybe in two years."
And it sounds responsible. Why would you build models on bad data?
Here's the reality: no organisation has perfect data. Not one. The organisations successfully using AI in asset management didn't wait for perfection — they started with what they had and improved the data as part of the AI project, not before it.
What we've found is:
→ You need good data on your critical assets — not all assets
→ You need consistent failure coding on target failure modes — not the entire CMMS
→ You need 2-3 years of history — not decades
→ You need one person who owns data quality — not a transformation programme
The "we're not ready" story feels safe. But it's often a way to avoid the harder conversation: "we don't know where to start."
Start small. Pick one critical asset. Clean that data. Build one model. Learn.
What's the myth about AI in your industry that you wish would go away?
A water utility was replacing pumps every 7 years — whether they needed it or not.
The result? $2.4M a year on pump maintenance. A third of the replacements were on pumps with years of life left. And the pumps that actually needed attention? They were failing between scheduled replacements, causing service disruptions.
Today, that same utility replaces pumps based on condition. Unplanned pump failures are down 68%. Maintenance spend dropped by $900K in the first year. And their operators trust the system because they helped build it.
Here's what bridged the gap:
→ They started with their 15 most critical pump stations — not the full fleet
→ They installed vibration sensors and connected them to their existing SCADA system
→ A reliability engineer mapped the dominant failure modes for each pump type
→ They built simple condition-based rules first, then layered in predictive models once the team trusted the data
→ The maintenance planners were involved from day one — not handed a dashboard after the fact
The lesson most people miss: the technology was the easy part. The hard part was getting the maintenance team to stop doing what they'd always done. That only happened because they were part of the design, not just the rollout.
Has your organisation made a shift like this — from calendar-based to condition-based? What was the hardest part of the transition?
The skill responds to these in-session commands:
| Command | Action |
|---|---|
interview | Run the mandatory interview to gather real material |
produce-all | Generate all eligible formats in parallel, review, and recommend the best |
draft | Generate a single post from the current brief in a specific format |
hooks | Generate 3–5 alternative hook options for the current topic |
shorter | Rewrite the current draft in fewer words |
longer | Expand the current draft with more detail |
spicier | Make the current draft more provocative or opinionated |
softer | Tone down the current draft — less confrontational |
carousel | Convert the current topic into carousel slide text |
slay | Rewrite the current topic as a SLAY story-led post |
confess | Rewrite the current topic as a confession post |
contrast | Rewrite the current topic as a This, Not That comparison |
myth | Rewrite the current topic as a Myth vs Reality debunking |
bab | Rewrite the current topic as a BABLA transformation story |
variations | Generate 2–3 different angles on the same topic |
checklist | Run the quality checklist against the current draft |
image | Generate a matching image for the current post using nano-banana-2 (1:1 aspect ratio) |
When in doubt, ask: "Would a senior asset management professional find this valuable and credible?" If yes, ship it. If not, revise.
npx claudepluginhub sas-asset-management/sasamclaudecodeskills --plugin linkedin-post-generatorGenerates two viral LinkedIn posts in proven formats, matched to the founder's voice, using battle-tested templates. Auto-selects formats based on topic.
Generates LinkedIn posts, carousels, newsletters, and 30-day calendars with niche-specific SEO and a reinforcement-learning memory system that adapts to your feedback.
Guides step-by-step repurposing of YouTube videos, blogs, guides, or insights into LinkedIn posts matching Ben's voice. Activates on LinkedIn post requests.