write-as-human
Skill4 starsChanged 3 months agoMITView on GitHub
Strip AI writing patterns from prose so it reads as human-written.
In this repository53 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- actionableSkill
- astro-bootstrapSkill
- astro-patternsSkill
- autopilotSkill
- axon-code-intelSkill
- brainstormingSkill
- commit-conventionsSkill
- cross-repo-contextSkill
- django-bootstrapSkill
- django-patternsSkill
- executeSkill
- find-sessionSkill
- flutter-bootstrapSkill
- flutter-patternsSkill
- full-featureSkill
- intelSkill
- nextjs-bootstrapSkill
- nextjs-fullstack-bootstrapSkill
- nextjs-fullstack-patternsSkill
- nextjs-mui-bootstrapSkill
- nextjs-patternsSkill
- nextjs-shadcn-bootstrapSkill
- orchestratorSkill
- phase-breakdownSkill
- plan-and-scopeSkill
- prdSkill
- reactSkill
- recallSkill
- reflectSkill
- relaySkill
- rememberSkill
- scaffoldSkill
- scratchSkill
- tailwindSkill
- ux-specSkill
- ux-to-promptsSkill
- verify-astroSkill
- verify-djangoSkill
- verify-flutterSkill
- verify-nextjs-fullstackSkill
- verify-nextjs-muiSkill
- verify-nextjs-shadcnSkill
- verify-nextjsSkill
- verifySkill
- visual-feedbackSkill
- work-disciplineSkill
- worktreesSkill
- wrapSkill
- write-as-friendSkill
- writing-clearlySkill
- writing-skillsSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 80
After writing, scan for these high-signal tells (the patterns readers catch fastest):
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Write as Human
Condensed from tropes.fyi by ossama.is
Overview
A checklist of AI writing tells to eliminate from prose. Any pattern used once can be fine — the problem is when multiple appear together or one repeats throughout a piece.
When to Use
- Writing or editing documentation, blog posts, READMEs, or long-form prose
- Reviewing AI-generated copy before publishing
- Any time prose "sounds like AI" and you need to diagnose why
Word Choice
| Pattern | What it looks like | Fix |
| Magic adverbs | "quietly", "deeply", "fundamentally", "remarkably" adding false weight | Cut the adverb or replace with something specific |
| Delve family | "delve", "utilize", "leverage", "harness", "streamline", "robust" | Use plain verbs: explore, use, take advantage of, strong |
| Ornate nouns | "tapestry", "landscape", "paradigm", "ecosystem" for any domain/field | Name the actual thing |
| "Serves as" dodge | "serves as", "stands as", "represents" replacing "is" | Just say "is" |
Sentence Structure
| Pattern | What it looks like | Fix |
| Negative parallelism | "It's not X — it's Y." False profundity via reframe. One per piece max. | State the point directly |
| Dramatic countdown | "Not X. Not Y. Just Z." | Lead with Z |
| Self-answered question | "The result? Devastating." Asking then immediately answering. | Merge into a single declarative sentence |
| Anaphora abuse | "They could... They could... They could..." Same opener 3+ times | Vary sentence openings |
| Tricolon abuse | Back-to-back rule-of-three constructions | One tricolon per section, max |
| Filler transitions | "It's worth noting", "Importantly", "Interestingly", "Notably" | Cut entirely — just make the point |
| Shallow -ing analysis | "...highlighting its importance", "...reflecting broader trends" | Either make the analysis specific or remove it |
| False ranges | "From X to Y" where X and Y aren't on a real scale | List the items plainly |
Paragraph and Document Structure
| Pattern | What it looks like | Fix |
| Punchy fragments | One-sentence paragraphs for manufactured emphasis | Combine into real paragraphs with developed thoughts |
| Listicle in prose | "The first... The second... The third..." disguised as paragraphs | Use an actual list, or write actual paragraphs |
| Fractal summaries | Intro summarizes, each section summarizes, conclusion re-summarizes | One summary, at most. Trust the reader |
| Dead metaphor | Same metaphor repeated 5+ times across a piece | Use it once, then move on |
| Historical stacking | "Apple didn't... Facebook didn't... Stripe didn't..." rapid-fire analogies | One comparison, well-developed |
| One-point dilution | Single argument restated 8 ways across 4000 words | State it once, support with evidence, stop |
| Signposted conclusion | "In conclusion", "To sum up", "In summary" | The reader knows it's the ending. Just end. |
Tone
| Pattern | What it looks like | Fix |
| False suspense | "Here's the kicker", "Here's the thing", "Here's where it gets interesting" | Drop the buildup, deliver the point |
| Patronizing analogy | "Think of it as a highway for data" | Explain directly; only use analogy if genuinely clarifying |
| "Imagine a world" | Futurist invitation followed by a wish list | Ground claims in present reality |
| False vulnerability | Performative self-awareness: "And yes, I'm openly..." | Real honesty is specific and uncomfortable. Skip the performance. |
| "The truth is simple" | Asserting clarity instead of demonstrating it | Prove the point; don't announce it |
| Stakes inflation | "This will fundamentally reshape everything" | Scope claims to what you can actually demonstrate |
| "Let's break this down" | Pedagogical hand-holding: "Let's unpack", "Let's explore" | Just start explaining |
| Vague attributions | "Experts argue", "Industry reports suggest" without naming anyone | Name the source or drop the claim |
| Invented concept labels | Compound terms like "supervision paradox", "acceleration trap" used as if established | Define it properly or don't coin it |
Formatting
| Pattern | What it looks like | Fix |
| Em-dash addiction | 10+ em dashes in a piece for drama and asides | 2-3 per piece, where natural |
| Bold-first bullets | Every list item starts with a bolded keyword | Use bold sparingly, only when scanning matters |
| Unicode decoration | Arrows (->), smart quotes, special characters | Use standard ASCII characters |
| "Despite its challenges" | Acknowledge problem, immediately dismiss it with optimism | Either engage with the problem or don't raise it |
Quick Self-Check
After writing, scan for these high-signal tells (the patterns readers catch fastest):
- Negative parallelism — Count "not X — it's Y" constructions. More than one is suspect.
- Bold-first bullets — If every list item starts bold, restructure.
- Em-dash count — More than 3 in a piece? Trim.
- Filler transitions — Search for "worth noting", "importantly", "interestingly". Delete all.
- Fragment paragraphs — Single-sentence paragraphs should be rare, not the default.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.