Claude Code plugin

deflate πŸ’©

Paste any LinkedIn thought-leadership post. Get back what actually happened β€” plus a bullshit meter, a brutal one-liner, and the manipulation tricks, named.

Install in 2 commands Try it in Claude ✨ GitHub β†’ Feeling a little evil? 😈

See it work

An invented-but-familiar specimen: a founder announces "10x faster onboarding" over a giant graphic, with zero methodology, tagged "part 3/7".

claude

> /bullshit:deflate [pasted post]

Shit-o-meter: πŸ’© 90% [β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘]

We say our onboarding is 10x faster. I didn't explain compared to what, or how we measured it.

"I have a big number, a giant graphic repeating it, and zero idea how I calculated it."

The tells: the entire post defends the number without ever saying how it was measured; a giant graphic with the hero number as a substitute for data; and "part 3/7" β€” content-calendar farming, this is a franchise now.

Install

Works anywhere Claude Code runs. Output always matches the post's language β€” Spanish in, Spanish out.

Get Claude Code

If you don't have it yet: npm install -g @anthropic-ai/claude-code, then run claude and log in.

Add the marketplace & install

Inside any Claude Code session:

/plugin marketplace add idlechara/bullshit
/plugin install bullshit@bullshit

Deflate

Paste a post as text β€” or drop in a screenshot:

/bullshit:deflate <the post>

The tells it catches

Every "tell" maps to a documented persuasion technique β€” not an invention.

Borrowed authority

"When I was at Google…" β€” a logo doing the arguing.

appeal to authority Β· Cialdini

Fake paradox

An obvious consequence dressed as a mind-bending contradiction.

curiosity gap Β· pseudo-profundity

Personal-epiphany arc

A mundane decision told as a hero's journey.

narrative transportation Β· Green & Brock

Bumper-sticker lesson

The post exists to deliver a fortune cookie.

processing fluency Β· unfalsifiability

Comment-farming question

"Agree? πŸ‘‡" β€” a lever on the ranking algorithm, not a question.

engagement bait Β· defined by Meta, 2017

Buzzword bolted on

"AI" sprinkled on content that would be identical without it.

trend-hijacking Β· status display

Cadence-theater

Every. Line. Its. Own. Paragraph.

fluency-as-truth Β· the "see more" fold

Giant hero number

A huge "10x" with no baseline, denominator, or method.

anchoring Β· Tversky & Kahneman

Preemptive defense

"This isn't about layoffs" β€” so it's about layoffs.

apophasis Β· classical rhetoric

Our users say

Real feedback, lightly translated from the original Chilean Spanish.

"It's great for summarizing corporate events β€” the typical HR stuff, managers, blah blah blah, wall of text lol."

β€” a user, deflating internal comms

"It wasn't anxiety. It was the daily standup."

β€” a user, after the diagnosis

"It helped me summarize and clean up the performance-review messages and all that stuff."

β€” a user, surviving review season
Send feedback πŸ’Œ

The serious part

The humor comes strictly from deflation, never invention β€” the skill only mocks what is literally on the page, and never adds facts, motives, or events the post doesn't state.

The framing owes everything to Harry Frankfurt's On Bullshit: these posts usually aren't lies β€” they're indifferent to truth, with every word optimized for engagement instead of accuracy. That's what makes them deflatable: there's always a smaller, truer sentence underneath.

Each tell is grounded in real research. The sources, so you can read the actual documents:

The full catalog, with the mechanism and a practical "how to spot it" test for each tell, ships inside the plugin: references/catalog.md. Ask the skill why a tell works and it answers from there β€” sources cited, roast off.