Removes signs of AI-generated writing to make text sound natural and human. Use when editing or reviewing text for AI patterns, humanizing content, or making writing sound less robotic. Detects 24 patterns including inflated symbolism, promotional language, em dash overuse, rule of three, and AI vocabulary words.
Make writing sound human without changing its meaning. Generic writing can fail at two layers: wording (patterns 1-24 below) and construction, i.e. how the piece is built (patterns 25-38). Use these as review prompts, not a checklist the text must obey.
reference/construction-patterns.md and check patterns 25-38. Make only grounded structural edits before polishing wording; do not restructure a piece to satisfy the checklist.These patterns are editorial heuristics for making writing better, not a detector. Never use them to claim a text was AI-written.
There is no single structure that makes prose human. When author samples exist:
When the target is Lachlan's voice, use speak-like-lachlan as the author guide. Author-specific guidance and same-mode examples override generic patterns in this skill.
Do not add direct reader address, time jumps, references, anecdotes, jokes, or ambivalence merely to make the text look human. If a fix needs a fact, opinion, or memory that is not present, ask; use a placeholder only when the user wants a draft with gaps marked.
Removing generic patterns is only half the job. Recover what the author actually thinks and notices:
1. Undue emphasis on significance/legacy Watch for: stands/serves as, testament, pivotal moment, crucial role, underscores importance, reflects broader, symbolizing, marking/shaping, evolving landscape
2. Undue emphasis on notability Watch for: independent coverage, media outlets, active social media presence
3. Superficial -ing analyses Watch for: highlighting, emphasizing, reflecting, symbolizing, showcasing, fostering
4. Promotional language Watch for: boasts, vibrant, rich, profound, nestled, groundbreaking, renowned, breathtaking, stunning
5. Vague attributions Watch for: Experts believe, Some critics argue, Industry reports, Observers have cited
6. Formulaic challenges sections Watch for: Despite challenges... continues to thrive, Future Outlook
7. Stock vocabulary clusters Watch for repeated or unnecessary use of: Additionally, delve, crucial, enhance, fostering, interplay, intricate, landscape (abstract), pivotal, showcase, tapestry, testament, underscore, vibrant. A listed word is not, by itself, a reason to edit it.
8. Copula avoidance Watch for: serves as, stands as, boasts, features, offers (instead of is/are/has)
9. Negative parallelisms Watch for: Not only...but..., It's not just X, it's Y
10. Rule of three overuse
11. Synonym cycling
12. False ranges
13. Dash overuse Watch for repeated em dashes doing the work of commas, periods, colons, and parentheses. Preserve dashes when they fit the author's normal style and improve the sentence.
14. Boldface overuse Remove mechanical bold emphasis on terms.
15. Inline-header vertical lists
16. Title Case in headings
17. Emojis Remove decorative 🚀 💡 ✅ and similar flourishes when they are not characteristic of the author or format. Preserve intentional emoji use.
18. Typography drift Do not change quote style merely to make text look more human. Preserve the typography expected by the author, format, or publishing system.
19. Chatbot artifacts Remove generic service-language such as "I hope this helps!", "Of course!", "Certainly!", or "Would you like..." when it is not a genuine part of the message. Do not ban ordinary closings the author actually uses.
20. Knowledge-cutoff disclaimers Remove boilerplate such as "While specific details are limited..." or "based on available information..." when it contributes nothing. Preserve dates and caveats that genuinely bound the claim.
21. Sycophantic tone Remove unearned praise such as "Great question!", "You're absolutely right!", or "That's an excellent point!" Preserve sincere acknowledgement when it serves the conversation.
22. Filler phrases
23. Excessive hedging
24. Generic positive conclusions
Construction is how the piece is built, not how it is worded. For definitions, evidence levels, and fixes, read reference/construction-patterns.md. It distinguishes research findings from nonfiction adaptations and broader editorial heuristics.
Never inject synthetic humanity: fixes must surface something true. Ask when only the author knows; use an explicit placeholder only when the user wants a draft with gaps marked.
If an AI detector is used, compare matched text boundaries, formats, and window-level results. Treat its output as one diagnostic signal, never the editing objective or evidence of authorship. Do not add artefacts to chase a score.
Surface patterns based on Wikipedia:Signs of AI writing. Construction-pattern evidence and provenance are documented in reference/construction-patterns.md.
Google Workspace CLI for Gmail, Calendar, Drive, Contacts, Sheets, and Docs.
Manage Apple Notes via the `memo` CLI on macOS (create, view, edit, delete, search, move, and export notes). Use when a user asks OpenClaw to add a note, list notes, search notes, or manage note folders.
Work with Obsidian vaults (plain Markdown notes) and automate via obsidian-cli.
Use when you need to control Slack from OpenClaw via the slack tool, including reacting to messages or pinning/unpinning items in Slack channels or DMs.
Manage Apple Reminders via remindctl CLI (list, add, edit, complete, delete). Supports lists, date filters, and JSON/plain output.
Manage Trello boards, lists, and cards via the Trello REST API.
Category:productivity