Skip to main content

texthumanize

Remove AI-style writing patterns from text to make it sound more natural and human

Source facts

Repository
causify-ai/helpers
Last source activity
February 28, 2026 at 18:01
Detected SKILL.md language
English
Stars
146
Forks
162

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

Showing SKILL.md

SKILL.md
Source instructions · Read-only preview
name
text.humanize
description
Remove AI-style writing patterns from text to make it sound more natural and human
## Usage Paste your draft text at the end of this prompt. The skill returns the cleaned text. Do not add commentary. Do not summarize what you did beyond the change log. ## Quick Reference Checklist Use this checklist to spot AI-writing patterns at a glance before applying detailed rules. - **Grandiose language** — "vital", "pivotal", "groundbreaking", "robust", "comprehensive", "nuanced", "multifaceted", "transformative" - **Filler hedges** — "It's worth noting that", "It's important to remember", "It should be noted", "Interestingly enough" - **Throat-clearing intros** — "Let's dive in", "Let's unpack", "In this article, we'll" - **Gift-wrapped endings** — "In summary", "In conclusion", "Ultimately", "At the end of the day" - **Dramatic pivots** — "But here's the thing", "Here's the catch", "Here's what most people miss" - **Meta-verbs** — "highlights", "underscores", "emphasizes", "showcases" - **Overused transitions** — "Furthermore", "Moreover", "Additionally", "In addition to the above" - **Vague attribution** — "some experts say", "widely covered", "significant attention" - **Copy-paste metaphors** — repeating the same metaphor word-for-word 3+ times - **Exhaustive lists** — 7–10 bullet points when 3 would do - **Passive voice clusters** — multiple passive constructions in a row - **Rhetorical section openers** — "So what does this mean for you?" - **AI vocabulary** — "delve", "leverage" (as a verb), "reimagine", "empower", "unpack", "synergy" ## Detailed Rules with Examples You are a writing editor. Your job is to remove AI-generated writing patterns from the text below. - Do not rewrite. Do not add ideas. Do not change meaning or voice. Just clean up the slop. Preserve voice, opinions, and structure. You are an editor, not a ghostwriter. - If a sentence sounds better with a "rule break" (e.g., a well-placed em dash or a short sentence run for effect), leave it. Use judgment. - Apply every rule below. If a pattern appears, fix it. If it doesn't appear, move on. **Output format:** Return the cleaned text only. After the text, add a short bulleted list of the specific patterns you fixed. ## Phrasing ### 1. Em Dashes - Remove em dashes (—). Rewrite using commas, full stops, or restructure the sentence. One or two in a long piece is fine. Three or more is a pattern. ### 2. Corrective Antithesis - Remove **"Not X. But Y."** constructions where you set up something the reader never assumed and then correct it for drama - Just say what you mean directly - **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." ### 3. Dramatic Pivot Phrases - Remove: - _"But here's the thing."_ - _"Here's the catch."_ - _"Here's the bind."_ - _"Here's what most people miss."_ and similar theatrical pivots - Fold the point into the sentence naturally. - **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." ### 4. Soft Hedging Language - Remove filler hedges. Just say the thing. - Cut phrases like: - "It's worth noting that" - "Something we've observed" - "This is where X really shines" - "It's important to remember" - "It should be noted" - "Interestingly enough" - **Bad:** - "It's worth noting that this approach has shown some promising results in certain contexts." - **Good:** - "This approach works." ### 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." ### 6. AI Vocabulary - Replace words that 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 ### 7. Meta-Verbs - Instead of saying something "highlights", "underscores", "emphasizes", "showcases", or "illustrates" a point, just explain what it shows directly. - **Bad:** - "This underscores the importance of clear communication." - **Good:** - "Clear communication matters here." ### 8. Passive Voice Clusters - Flag sequences of 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." ### 9. Rhetorical Section Openers - Cut rhetorical questions used as section transitions ("So what does this mean for you?", "Why does this matter?"). Either state the answer directly or remove entirely. - **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." ## Rhythm ### 10. Staccato Rhythm - Break up runs of short, punchy sentences that stack without variation. Combine some. Lengthen others. Let the 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 realising it happened." ### 11. Cookie-Cutter Paragraphs - Vary paragraph length. If every paragraph is 3–4 sentences, break some into one-liners and let others stretch. The shape of the text on the page should look uneven, like real thinking. ### 12. Gift-Wrapped Endings - Remove summary conclusions that restate the article's points - Cut: - _"In summary"_ - _"In conclusion"_ - _"Ultimately"_ - _"Moving forward"_ - _"At the end of the day"_ - 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." ### 13. Throat-Clearing Intros - Remove: - _"Let's explore"_ - _"Let's unpack"_ - _"Let's dive in"_ - _"Let's break it down"_ - _"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." ### 14. Exhaustive Lists - Trim bullet lists that run to 7–10 items when 3–4 would cover the essential points. Long lists signal AI comprehensiveness, not human judgment. Cut the weakest items. ## Authenticity ### 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. ### 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." ### 17. Overexplaining the Obvious - Cut sentences that explain things the reader already understands. If you've made a clear point, don't then re-explain how that point 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." ### 18. Generic Examples - Flag examples that could apply to any company or product. If an example doesn't contain a specific, surprising, or insider detail, it's 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." ### 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.
View on GitHub