Guide
Best Claude Code skills for writing
Left alone, an agent writes in a recognisable way. A writing skill names the habits to cut and keeps the facts fixed. Here is the free one with its real rules and a before and after taken from the skill itself, then the paid options for voice and social posts.
Start free: Humanizer
Humanizer
FreeScan 100/100Strips the most common AI tells. Loads when the user writes or edits any prose. About 1,024 tokens when it triggers, 104 always loaded.
Use when the user writes, edits or pastes prose and says \"humanize this\", \"make it sound less like AI\", \"too robotic\", or before you hand over any post, email, landing copy or doc you drafted. Strips the most common AI tells: banned words, em-dash asides, not-X-but-Y framing, signposts and summary endings, in one pass without changing facts. Reads profile.md for brand, voice and never-say words.## Structural tells to cut
- "Not X, but Y" and "it's not just X, it's Y": state Y.
- Em dashes used for asides: use a comma, parentheses or a new sentence. Target zero.
- Signposts: "Let's dive in", "Here's the thing", "Here's why that matters".
- A question you then answer ("The result? Faster deploys.").
- Colon reveals ("The secret: consistency.").
- Summary endings: "In conclusion", "Ultimately", "At the end of the day", or a last paragraph that repeats the post.
- Lists of exactly three adjectives or clauses when fewer or more exist.## Worked example
Before: "We're thrilled to announce that our new dashboard isn't just a redesign, it's a game-changer. Seamless integrations empower teams to unlock insights. Ready to dive in?"
After (Direct): "The new dashboard is live. It connects to your existing tools and shows each metric next to the rows behind it. Open it from the sidebar."
Cut: the not-just frame, three word tells, the rhetorical close. If the source never named the integrations, keep the claim general and flag it to the user instead of inventing specifics.The example above is part of the skill's own text. It shows the shape of the edit: the not-just frame, the stock words and the rhetorical close go, and every fact stays.
Free for social posts: Hook Lab
Hook Lab
FreeScan 100/100Five hooks per post, each a different pattern. Loads when the user writes the first line of a post. About 971 tokens when it triggers, 87 always loaded.
Use when the user writes or rewrites the first line of a post, tweet, LinkedIn update, caption or video opener, or says \"I need a hook\", \"first line\", \"this post is not landing\". Writes five hooks per post, each a different pattern, each under twelve words and a specific claim. Reads profile.md for brand, audience and platforms.## Patterns
| Pattern | Shape |
|---|---|
| Result | The outcome, with the number: "Cut onboarding from 9 steps to 3." |
| Contrarian | A claim against the common belief, with the reason. |
| Mistake | What you did wrong and what it cost, in the first person. |
| Process | "How X does Y in N steps", with the real N. |
| Before and after | Two states, one line. |
| Named enemy | A specific thing the audience hates, called out. |
| Quote | A line someone actually said, with who. |
| Time stamp | A precise moment: "At 2am on a Tuesday, the deploy failed." |Paid: voice and tone
- Humanizer Pro$9Scan 100/1005 rules · HumanizerThree-pass rewrite. Facts locked. Your voice.Loads when the user writes, edits or humanizes any prose.
- Tone Lint$7Scan 100/1003 rules · HumanizerCatches drift before you publish.Loads when the user reviews a draft.
Use these when the free pass is not enough: a draft that must sound like one person, or a review that catches drift before you publish.
Paid: social formats
- Thread Builder$7Scan 100/1003 rules · Viral HooksThreads that hold attention.Loads when the user writes a thread.
- Carousel Writer$9Scan 100/1003 rules · Viral HooksSlide-by-slide copy and layout.Loads when the user makes a carousel.
- Content Engine$12Scan 100/1005 rules · Viral HooksOne idea, a week of native posts.Loads when the user wants content from one idea.
Which one first
| You want to… | Use |
|---|---|
| Clean up a draft you already have | Humanizer |
| Rewrite in your own voice from samples | Humanizer Pro |
| Catch tone drift before publishing | Tone Lint |
| Write a stronger first line | Hook Lab |
| Turn one idea into a week of posts | Content Engine |
Writing outreach instead? See skills for sales outreach. Planning a build? See skills for planning.
Questions
- What is a humanizer skill?
- A skill that tells the agent which habits make prose read as machine-written, such as signposts, not-X-but-Y framing and stock words, and to remove them without changing the facts. It is a clean-up pass, not a new draft.
- Does a humanizer skill make text undetectable?
- No, and none can promise that. It removes common stylistic tells so the text reads better to a person. Do not use any tool to claim machine-written text as human work where that matters.
- Can a skill match my voice?
- Partly. The free Humanizer takes a voice setting and a list of words you never use. The paid Humanizer Pro adds a pass built around samples of your own writing. Neither replaces reading the result yourself.
- Where does the skill read my settings?
- From a profile.md file in the skill folder, which you edit. The skill asks once for anything missing and writes the answer there.
Keep reading
- Skill ScannerScore any skill in your browser. Nothing is uploaded.
- SKILL.md format referenceEvery field, its limit, the folder layout and the errors people hit.
- Best Claude Code skills for sales outreachResearch, openers and follow-ups, with guardrails against spam.
- Best Claude Code skills for planningA free questioning skill, decision records and scope cutting.
- All guidesEvery guide, grouped by what you are trying to do.
Paste a SKILL.md or drop a folder. It is scored in your browser and nothing is uploaded.
Want a skill that already passes? Start with a free one.
HumanizerFreeScan 100/100Strips the most common AI tells. From the Humanizer collection.