home / skills / cdeistopened / opened-vault / ai-tells

ai-tells skill

safe

This skill helps you write authentic, human-sounding content by avoiding common AI tells, delivering clear, direct messaging across newsletters, blogs, and

npx playbooks add skill cdeistopened/opened-vault --skill ai-tells

Review the files below or copy the command above to add this skill to your agents.

Files (7)

SKILL.md

1.6 KB

---
name: ai-tells
description: Patterns that make writing sound AI-generated. Invoke this skill for ANY writing task - newsletters, social media, blog posts, emails. Apply these constraints automatically.
---

# AI Tells

Patterns to avoid in all written content. Read the examples in `references/` for voice - don't follow rules mechanically.

## The One Rule

**Never use the correlative construction.** This is the #1 AI tell:
- "X isn't just Y - it's Z"
- "It's not about X, it's about Y"
- "Schools aren't just buildings - they're communities"

Find another way. State the thing directly.

## Other Patterns to Avoid

1. **Setup phrases:** "The best part? ..." / "The secret? ..." / "Here's the thing..."
2. **Staccato fragments:** "No fluff. No filler. Just results."
3. **Triple adjectives:** "Fast, efficient, reliable" (grouping in threes)
4. **Cliché openers:** "In today's fast-paced..." / "In the ever-evolving..."
5. **Dramatic punctuation:** "Let that sink in" / "Now more than ever"
6. **Thesaurus abuse:** utilize, implement, facilitate, leverage, comprehensive, crucial
7. **Fake questions:** "Sound familiar?" / "Ready to level up?"
8. **Transition tells:** "And here's the kicker" / "Enter: [thing]"

## Format Constraints

- **Dashes:** hyphens with spaces - like this - not em dashes (—)
- **No emojis** in body content
- **No bold** for emphasis within paragraphs

## Voice References

Read these for the vibe, not as rules:
- `references/pirate-wires-examples.md` - 11 annotated examples showing voice in action
- `references/pirate-wires-style.md` - Voice breakdown

**The test:** Does this sound like a person talking, or a person performing "good writing"?

Overview

This skill detects and removes common patterns that make writing sound AI-generated. It applies a clear, human-first filter to any writing task - newsletters, social posts, blog articles, or emails. Use it to produce natural, readable copy that sounds like a real person wrote it.

How this skill works

The skill scans text for specific patterns: correlative constructions, setup phrases, staccato fragments, triple lists, clichéd openers, dramatic punctuation, overused synonyms, fake questions, and certain transition tells. It suggests alternate phrasings and enforces format constraints such as hyphens with spaces and no emojis. The output rewrites flagged segments to preserve meaning while restoring authentic voice.

When to use it

Before publishing newsletters or marketing emails
When preparing social media captions or threads
When editing blog posts for a human tone
To polish internal communications and reports
During final pass on client-facing copy

Best practices

Run the skill early, then iterate on voice and detail
Prefer direct statements over correlative constructions
Avoid filler setup phrases and fake rhetorical questions
Keep lists balanced - use two items or full sentences instead of triple adjectives
Follow the format rules: hyphens with spaces, no emojis

Example use cases

Turn a product announcement into a concise, human message
Edit a weekly newsletter so it reads like a colleague wrote it
Convert a draft blog post to a natural narrative voice
Refine social captions to avoid buzzword-heavy phrasing
Clean up customer-facing emails for clarity and warmth

FAQ

Will the skill change the meaning of my text?

No. Changes aim to preserve meaning while shifting phrasing toward natural language and removing flagged patterns.

Does it block certain words?

It flags and suggests alternatives for overused synonyms and canned phrases rather than outright blocking them.