gpsaggese/helpers / .claude/skills/text.humanize
text.humanize
VerifiedRemove AI-style writing patterns from text to sound more natural and human
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.3 KB |
SKILL.md
View on GitHub---
description: Remove AI-style writing patterns from text to sound more natural and human
model: haiku
---
# Goal
- Change text with recognizable AI mannerisms (hedging, dramatic pivots, cliched
transitions). Clean up AI-written content that is factually correct but flat
# What This Skill Preserves
- Author's opinions, arguments, and claims
- Document structure and organization
- Paragraph and section order
- Core meaning and facts
- Intentional stylistic choices (fragments, unconventional punctuation, etc.)
# Workflow
## Core Principles
- Edit, don't rewrite. Clean up pattern slop, keep the voice, opinions, structure,
and meaning
- Apply systematically: check all 19 rules below. Skip a rule only if it does not
apply
- Use judgment on conflicts. If breaking a rule serves the writing (e.g., a
well-placed em dash for effect), keep it
- Change nothing else: no reordered paragraphs, added ideas, rephrased core
arguments, or altered facts
## Pattern Checklist
Scan for these AI markers before applying detailed rules:
- Overused Words and Jargon
- **Grandiose**: "vital", "pivotal", "groundbreaking", "robust", "comprehensive",
"nuanced", "multifaceted", "transformative"
- **AI vocabulary**: "delve", "leverage" (verb), "reimagine", "empower", "unpack",
"synergy"
- Filler and Hedges
- "It's worth noting that", "It's important to remember", "It should be
noted", "Interestingly enough"
- "Something we've observed", "This is where X really shines"
- Theatrical Transitions and Pivots
- **Throat-clearing**: "Let's dive in", "Let's unpack", "In this article, we'll"
- **Dramatic pivots**: "But here's the thing", "Here's the catch", "Here's what
most people miss"
- **Overused connectors**: "Furthermore", "Moreover", "Additionally", "In addition
to the above"
- **Rhetorical openers**: "So what does this mean for you?", "Why does this
matter?"
- Structural Patterns
- **Gift-wrapped endings**: "In summary", "In conclusion", "Ultimately", "At the
end of the day", "Moving forward"
- **Exhaustive lists**: 7-10 items when 3-4 would suffice
- **Passive voice clusters**: 2+ passive sentences in a row
- **Corrective antithesis**: "Not X. But Y." setup-payoff constructions
- Meta-Language and Vagueness
- **Meta-verbs**: "highlights", "underscores", "emphasizes", "showcases",
"illustrates"
- **Vague attribution**: "some experts say", "widely covered", "significant
attention"
- **Copy-paste metaphors**: same metaphor repeated word-for-word 3+ times
## Rules
### Rule 1: Em Dashes
- Remove excessive em dashes (—)
- Rewrite with commas, full stops, or a restructured sentence
- One or two in a long piece is fine. Three or more is a pattern worth fixing
### Rule 2: Corrective Antithesis
- Remove "Not X. But Y." constructions that set up something the reader never
assumed, then correct it for drama
- **Bad**: "This isn't because they don't trust the technology. It's because they
can't predict it."
- **Good**: "They trust the technology fine. What they can't do is predict it."
### Rule 3: Dramatic Pivot Phrases
- Remove theatrical pivots: "But here's the thing", "Here's the catch", "Here's
the bind", "Here's what most people miss"
- Fold the point into the sentence
- **Bad**: "The patterns are valuable. But here's the bind: building a tool cost
more than most could justify."
- **Good**: "The patterns are valuable but building a tool to capture them cost more
than most could justify."
### Rule 4: Soft Hedging Language
- Cut filler hedges. Just say the thing
- Remove the filler hedges listed in `## Pattern Checklist` (Filler and Hedges)
- **Bad**: "It's worth noting that this approach has shown some promising results in
certain contexts."
- **Good**: "This approach works."
### Rule 5: Overused Transition Words
- Cut or vary "Furthermore", "Moreover", "Additionally", "In addition to the
above" when chained together
- Real writers use them sparingly
- **Bad**: "The system is fast. Furthermore, it is reliable. Moreover, it is easy to
use. Additionally, it integrates well."
- **Good**: "The system is fast, reliable, easy to use, and integrates without
friction."
### Rule 6: AI Vocabulary
- Replace words AI overuses with plain alternatives
- "delve" -> explore, look at, examine
- "leverage" (verb) -> use, apply, rely on
- "robust" -> strong, solid, reliable
- "comprehensive" -> thorough, complete, full
- "nuanced" -> subtle, layered, specific
- "multifaceted" -> complex, varied
- "transformative" -> significant, major
- "unpack" -> explain, break down
- "reimagine" -> rethink, redesign
- "empower" -> let, help, enable
### Rule 7: Meta-Verbs
- Don't say something "highlights", "underscores", "emphasizes", "showcases", or
"illustrates" a point
- Explain what it actually shows
- **Bad**: "This underscores the importance of clear communication."
- **Good**: "Clear communication matters here."
### Rule 8: Passive Voice Clusters
- Flag two or more passive constructions in a row
- Rewrite at least one in active voice to restore momentum
- **Bad**: "The report was reviewed by the team. Errors were identified. Changes
were recommended."
- **Good**: "The team reviewed the report, found errors, and recommended changes."
### Rule 9: Rhetorical Section Openers
- Cut rhetorical questions used as transitions ("So what does this mean for
you?", "Why does this matter?")
- State the answer directly, or remove it
- **Bad**: "So what does this mean for your team? It means you need to rethink your
process."
- **Good**: "Your team needs to rethink its process."
### Rule 10: Staccato Rhythm
- Break up runs of short, punchy sentences that stack without variation
- Combine some. Lengthen others
- Let rhythm follow the thinking, not a drumbeat
- **Bad**: "Now, agents act. They send emails. They modify code. They book
appointments."
- **Good**: "Agents are starting to do real things now. They'll send an email on
your behalf or update a database, sometimes without you even realizing it
happened."
### Rule 11: Cookie-Cutter Paragraphs
- Vary paragraph length. If every paragraph runs 3-4 sentences, split some and
stretch others
- The page should look uneven, like real thinking
### Rule 12: Gift-Wrapped Endings
- Remove summary conclusions that restate the article's points
- **Cut** the closers listed in `## Pattern Checklist` (Gift-wrapped endings)
- End with something specific, human, or unresolved
- **Bad**: "In summary, by focusing on clear communication, consistent feedback, and
mutual trust, teams can build stronger relationships."
- **Good**: "The best teams I've worked with never talked about trust. They just had
it."
### Rule 13: Throat-Clearing Intros
- Remove throat-clearing openers such as "Let's explore", "Let's dive in", and
"In this article, we'll"
- Just start. The best first sentence puts the reader in the middle of something
- **Bad**: "In this article, we'll explore the hidden costs of micromanagement
Let's dive in."
- **Good**: "I micromanaged someone last Tuesday."
### Rule 14: Exhaustive Lists
- Trim bullet lists of 7-10 items to the 3-4 that cover the essential points
- Long lists signal AI comprehensiveness, not human judgment
- Cut the weakest items
### Rule 15: Perfect Punctuation
- Don't correct every grammar "mistake" if it sounds more natural broken
- Fragments are fine
- Starting with "And" or "But" is fine
- A comma splice can stay if it reads well
- If the draft has personality in its punctuation, keep it
### Rule 16: Copy-Paste Metaphors
- If the same metaphor or phrase appears more than twice, vary the language
- Use a pronoun, rephrase it, or trust the reader to remember
- Never repeat a metaphor word-for-word three times
- **Bad**: "Trust is like a battery. When the trust battery is full... But when the
trust battery runs low... To recharge the trust battery..."
- **Good**: "Trust is like a battery. When it's full, you barely think about it. But
let it drain and suddenly every interaction needs a charger."
### Rule 17: Overexplaining the Obvious
- Cut sentences that explain what the reader already understands
- If you've made a clear point, don't re-explain how it works
- **Bad**: "Trust is earned over time. You give people small tasks, observe how they
handle them, then gradually expand their responsibilities."
- **Good**: "Trust is earned. Everyone knows this. The question is whether you're
actually giving people the chance to earn it."
### Rule 18: Generic Examples
- Flag examples that could apply to any company or product
- An example without a specific, surprising, or insider detail is filler
- Either sharpen it or cut it
- **Bad**: "Take Slack, for example. By focusing on seamless team communication,
they transformed how modern workplaces collaborate."
- **Good**: "Slack solved the wrong problem brilliantly. Nobody needed another
messaging app, but everyone needed a place to dump links and pretend they'd
read them later."
### Rule 19: Vague Attribution
- Avoid vague references like "some experts say", "widely covered", "significant
attention"
- Identify the actual critic, report, study, or author when possible
- If you can't name the source, cut the claim
## Save
- Save the cleaned text over the input file. Change only the text, not the format
# Verification
- [ ] Confirm the file's structure, order, and meaning are unchanged
- [ ] Confirm no AI markers listed in `## Pattern Checklist` remain
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |