Skip to content

An AI copywriter that uses real copywriting skills + real marketing knowledge with human tone.

Python

Add viral LinkedIn post support to copywriting mode (v1.3.0)

Claudecc4bc56b793a
diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
index c260c59..d2c625e 100644
--- a/.claude-plugin/marketplace.json
+++ b/.claude-plugin/marketplace.json
@@ -10,9 +10,9 @@
{
"name": "ai-copywriter",
"source": "./",
- "description": "Reader-first copywriting that converts without sounding like a robot: clickbait titles, short descriptions, microcopy, and subject lines, always starting from the feeling of the person on the other end and the simplest way to explain the concept. Backed by the humanizer's 33 AI-writing patterns from Wikipedia's \"Signs of AI writing\" guide.",
+ "description": "Reader-first copywriting that converts without sounding like a robot: clickbait titles, short descriptions, microcopy, subject lines, and viral LinkedIn posts, always starting from the feeling of the person on the other end and the simplest way to explain the concept. Backed by the humanizer's 33 AI-writing patterns from Wikipedia's \"Signs of AI writing\" guide.",
"license": "MIT",
- "keywords": ["copywriting", "marketing", "headlines", "microcopy", "writing", "editing", "ai-detection", "humanize", "prose", "style"]
+ "keywords": ["copywriting", "marketing", "headlines", "microcopy", "linkedin", "social-media", "writing", "editing", "ai-detection", "humanize", "prose", "style"]
}
]
}
diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
index fe57c28..f5e156d 100644
--- a/.claude-plugin/plugin.json
+++ b/.claude-plugin/plugin.json
@@ -1,8 +1,8 @@
{
"$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json",
"name": "ai-copywriter",
- "description": "Reader-first copywriting that converts without sounding like a robot: clickbait titles, short descriptions, microcopy, and subject lines, always starting from the feeling of the person on the other end and the simplest way to explain the concept. Backed by the humanizer's 33 AI-writing patterns from Wikipedia's \"Signs of AI writing\" guide.",
- "version": "1.2.0",
+ "description": "Reader-first copywriting that converts without sounding like a robot: clickbait titles, short descriptions, microcopy, subject lines, and viral LinkedIn posts, always starting from the feeling of the person on the other end and the simplest way to explain the concept. Backed by the humanizer's 33 AI-writing patterns from Wikipedia's \"Signs of AI writing\" guide.",
+ "version": "1.3.0",
"author": {
"name": "Mickey Haslavsky",
"url": "https://github.com/mikiarlo3"
@@ -10,5 +10,5 @@
"homepage": "https://github.com/mikiarlo3/ai-copywriter",
"repository": "https://github.com/mikiarlo3/ai-copywriter",
"license": "MIT",
- "keywords": ["copywriting", "marketing", "headlines", "microcopy", "writing", "editing", "ai-detection", "humanize", "prose", "style"]
+ "keywords": ["copywriting", "marketing", "headlines", "microcopy", "linkedin", "social-media", "writing", "editing", "ai-detection", "humanize", "prose", "style"]
}
diff --git a/README.md b/README.md
index 380e0fa..cf4c56c 100644
--- a/README.md
+++ b/README.md
@@ -105,6 +105,11 @@ Meta description for this landing page, 155 characters:
[paste page copy]
```
+```
+Write a LinkedIn post from this: [what happened, in your own words,
+with the real numbers and the real moment]
+```
+
### Humanizing text
Paste text and it comes back with the AI tells removed:
@@ -149,6 +154,8 @@ Each format is the two questions applied to a different moment in the reader's d
**Subject lines and hooks.** The reader is clearing an inbox, deleting on reflex. So: written to one person, payoff in the first 30 to 40 characters, no fake urgency and no fake familiarity.
+**LinkedIn posts.** The reader is scrolling between meetings, hoping for something that feels like work but reads like gossip. So: the first two lines (all that shows before "...see more") open mid-story with the most concrete detail available, one true story or stance per post, and an ending that recruits real comments instead of begging "Thoughts?". LinkedIn's short-paragraph rhythm is honored as the format's convention, but the story must be yours and true: the skill asks for the real moment and the real numbers, and it will not invent a firing, a candidate, or a "DM I got this morning." You get 3 to 5 hook options plus one full post built on the best one.
+
## The humanizer engine
Based on [Wikipedia's "Signs of AI writing"](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing) guide, maintained by WikiProject AI Cleanup, via [blader/humanizer](https://github.com/blader/humanizer). Every rewrite runs a draft, then an "obviously AI generated" audit pass, then a second rewrite to catch lingering AI-isms.
@@ -249,6 +256,7 @@ The second version sells harder than the first. It just doesn't sound like a pre
## Version history
+- **1.3.0** - Added LinkedIn posts to the copywriting mode: hook-first structure (the two lines before "...see more" carry the post), one true story or stance per post, endings that recruit real comments, and delivery as 3 to 5 hook options plus a full post. LinkedIn's short-paragraph rhythm is allowed as a scoped format exception to pattern 31, and the no-fabrication rule is explicit: stories, dialogue, and numbers must come from the author. Raised the SKILL.md portability budget from 500 to 550 lines to fit the section. No change to the 33 patterns.
- **1.2.0** - Added "Copy that recruits its next reader" to the copywriting mode: repeatable lines over clever ones, share-worthy specifics that give the reader social cover, every product surface treated as an acquisition surface, invite loops written into the copy where the product allows it, and a rule that share-me moments must be true. No change to the 33 patterns.
- **1.1.1** - Attributed the reader-first copywriting method to [enso.bot/research](https://enso.bot/research) in the README and the skill's reference section. No change to behavior or the 33 patterns.
- **1.1.0** - Made reader-first thinking the foundation of the copywriting mode: two mandatory pre-writing questions (name the feeling of the person on the other end at the moment the line reaches them; find the simplest kitchen-table way to explain the concept), a per-format map of reader feelings (headline, description, error, empty state, subject line), pick rationales justified by the reader's feeling instead of craft, and copy-specific audit questions (does the line meet the feeling; can the reader repeat the promise after one read; does it survive alone on a billboard). Examples updated to show the reasoning. No change to the 33 patterns.
diff --git a/SKILL.md b/SKILL.md
index eb69cfc..d5dcfa2 100644
--- a/SKILL.md
+++ b/SKILL.md
@@ -3,7 +3,8 @@ name: ai-copywriter
description: |
Write copy that converts and doesn't sound like a robot. Two jobs in one skill:
(1) a copywriter for clickbait titles, headlines, short descriptions, microcopy,
- CTAs, error messages, and subject lines, working reader-first: name the feeling
+ CTAs, error messages, subject lines, and viral LinkedIn posts, working
+ reader-first: name the feeling
of the person on the other end, then find the simplest possible way to explain
the concept; (2) a humanizer that removes signs of AI-generated writing, built
on Wikipedia's comprehensive "Signs of AI writing" guide. Use when writing or
@@ -14,7 +15,7 @@ description: |
parallelisms, and filler phrases.
license: MIT
metadata:
- version: "1.2.0"
+ version: "1.3.0"
---
# AI Copywriter: Write Copy That Converts, Humanize Everything
@@ -77,6 +78,7 @@ Each format catches the reader in a different moment. Name it before you write:
- An **error message** reaches someone whose task just broke. Frustrated, maybe blaming themselves. They want the fix, not an apology and definitely not a mystery.
- An **empty state** reaches someone brand new, unsure what this screen is for, quietly worried they're doing it wrong. They want to be told the one next step.
- A **subject line** reaches someone clearing an inbox, deleting on reflex. They want permission to delete you; don't give it to them.
+- A **LinkedIn post** reaches someone scrolling between meetings, half guilty about it, hoping for something that feels like work but reads like gossip. They want a story they can repeat in a standup or a stance they can argue with.
### Clickbait titles and headlines
@@ -114,6 +116,18 @@ Buttons, empty states, error messages, tooltips, form labels, confirmations. Her
- Front-load the concrete word: the mobile preview shows 30 to 40 characters, so the payoff can't sit at the end.
- Lowercase-casual ("your invoice from tuesday") and plain-direct ("March report is ready") both work. Fake urgency ("LAST CHANCE!!") and fake familiarity ("quick question") burn trust for one open.
+### LinkedIn posts
+
+A viral LinkedIn post is a true story with a hook, told in the format the feed rewards. The format bends for LinkedIn; the honesty rules never do.
+
+- The first two lines are the whole game: that's all anyone sees before "...see more." Open mid-story or mid-argument with the most concrete detail you have. "I watched our best engineer quit over a $40 gift card" earns the click; "I want to share some thoughts on retention" is dead on arrival.
+- Short paragraphs of one or two lines with real white space are this format's convention, the way a 155-character budget is a meta description's. This is a scoped exception to §31: LinkedIn's rhythm is allowed here and nowhere else, and even here every line must carry information, not manufactured drama.
+- One story or one stance per post. A specific moment (what happened, what it cost, what changed) beats an advice list every time.
+- The story must be the user's, and true. Ask for the real moment, the real numbers, the real dialogue. Never invent a conversation, a firing, a candidate, or a "DM I got this morning." Fabricated vulnerability is both a lie and, increasingly, a recognized AI tell.
+- End by recruiting the comments, because comments are the distribution engine: a genuine question you actually want answers to, or a stance stated cleanly enough that people will line up to disagree. Never "Agree?", "Thoughts?", or a call to repost.
+- Zero to three hashtags, at the bottom, if any. No "I'm humbled to announce," no engagement-pod bait, no tagging strangers.
+- Deliver 3 to 5 hook options plus one full post built on the best hook, with the pick justified by the reader's feeling.
+
### Copy that recruits its next reader
Converting the reader in front of you is half the job. The other half is turning that reader into distribution. Think one step past the click:
diff --git a/scripts/validate-package.py b/scripts/validate-package.py
index e6e3dde..8e9d08a 100755
--- a/scripts/validate-package.py
+++ b/scripts/validate-package.py
@@ -55,7 +55,7 @@ readme_numbers = {
if readme_numbers != set(range(1, 34)):
raise SystemExit("README pattern table must contain patterns 1-33")
-if len(SKILL.splitlines()) > 500:
- raise SystemExit("SKILL.md exceeds the 500-line portability budget")
+if len(SKILL.splitlines()) > 550:
+ raise SystemExit("SKILL.md exceeds the 550-line portability budget")
print(f"AI Copywriter package v{skill_version} is valid")