Skip to content
Back to search

human-speak

Remove recognizable AI writing patterns without flattening the writer's voice.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/nateherkai--human-speak/badge.svg)](/listings/nateherkai--human-speak)

HTML

<a href="/listings/nateherkai--human-speak"><img src="/api/listings/nateherkai--human-speak/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

The simplest cross-platform installation uses the open skills installer:

npx skills add nateherkai/human-speak --skill human-speak --global --yes

You can also ask Claude or Codex:

Install the human-speak skill globally from https://github.com/nateherkai/human-speak

Claude Code plugin installation

/plugin marketplace add nateherkai/human-speak
/plugin install human-speak@human-speak

Claude Code namespaces plugin skills, so the plugin form is available as `/human-speak:human-speak`. Use the standalone skill installation above when you want the shorter `/human-speak` command.

Codex

After installation, invoke the skill as `$human-speak`. Codex uses `$skill-name` syntax rather than slash commands.

Updating an existing installation

GitHub updates do not automatically replace copies already installed on your computer. Reinstall using your original installation method, or ask your assistant:

Update my installed human-speak skill from https://github.com/nateherkai/human-speak to the latest version. Verify the installed SKILL.md, pattern library, and evaluation checklist against the repository.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 27 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "name": "human-speak",
  3. "description": "Writing tools that remove recognizable AI patterns without flattening the writer's voice.",
  4. "owner": {
  5. "name": "Nate Herk",
  6. "email": "253835467+nateherkai@users.noreply.github.com"
  7. },
  8. "plugins": [
  9. {
  10. "name": "human-speak",
  11. "displayName": "Human Speak",
  12. "source": "./",
  13. "description": "Make AI-assisted writing sound direct, specific, and human.",
  14. "author": {
  15. "name": "Nate Herk",
  16. "email": "253835467+nateherkai@users.noreply.github.com",
  17. "url": "https://github.com/nateherkai"
  18. },
  19. "homepage": "https://github.com/nateherkai/human-speak",
  20. "repository": "https://github.com/nateherkai/human-speak",
  21. "license": "MIT",
  22. "keywords": ["writing", "editing", "humanize", "ai-slop"],
  23. "category": "Productivity",
  24. "tags": ["writing", "editing", "voice"]
  25. }
  26. ]
  27. }
README.md · README · 127 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Human Speak
  2. Remove recognizable AI writing patterns without flattening the writer's voice.
  3. Human Speak edits, audits, and drafts writing that sounds direct, specific, and natural. It catches canned insight framing, fake reveals, dramatic fragments, vague authority, inflated buzzwords, reaction narration, recap endings, and other patterns that make AI-assisted writing sound interchangeable.
  4. It also protects the parts many editing prompts erase: vocabulary, cadence, bluntness, humor, uncertainty, digressions, and useful imperfections.
  5. Current version: **1.1.0**. See the [changes](CHANGELOG.md) and the [ranked 20-pattern checklist](skills/human-speak/references/top-20.md).
  6. ## Install the skill
  7. The simplest cross-platform installation uses the open skills installer:
  8. ```sh
  9. npx skills add nateherkai/human-speak --skill human-speak --global --yes
  10. ```
  11. You can also ask Claude or Codex:
  12. ```text
  13. Install the human-speak skill globally from https://github.com/nateherkai/human-speak
  14. ```
  15. ### Claude Code plugin installation
  16. ```text
  17. /plugin marketplace add nateherkai/human-speak
  18. /plugin install human-speak@human-speak
  19. ```
  20. Claude Code namespaces plugin skills, so the plugin form is available as `/human-speak:human-speak`. Use the standalone skill installation above when you want the shorter `/human-speak` command.
  21. ### Codex
  22. After installation, invoke the skill as `$human-speak`. Codex uses `$skill-name` syntax rather than slash commands.
  23. ### Updating an existing installation
  24. GitHub updates do not automatically replace copies already installed on your computer. Reinstall using your original installation method, or ask your assistant:
  25. ```text
  26. Update my installed human-speak skill from https://github.com/nateherkai/human-speak to the latest version. Verify the installed SKILL.md, pattern library, and evaluation checklist against the repository.
  27. ```
  28. ## Use it
  29. ### Edit a draft
  30. ```text
  31. /human-speak
  32. [paste draft]
  33. ```
  34. Human Speak returns the edited draft and a short explanation of what changed. Ask for "clean copy only" when you do not want the notes.
  35. ### Audit without rewriting
  36. ```text
  37. /human-speak audit this without rewriting:
  38. [paste draft]
  39. ```
  40. The audit names each pattern, quotes the evidence, and gives a short fix. It does not guess whether AI wrote the text.
  41. ### Draft new copy
  42. ```text
  43. /human-speak write a LinkedIn post from these notes in my voice:
  44. [paste notes]
  45. ```
  46. ## What it catches
  47. - Secret-insight framing such as "What nobody tells you"
  48. - Tee-ups such as "Here's where it gets crazy"
  49. - Binary reframes such as "It's not X. It's Y."
  50. - Mechanical three-beat rhythms and dramatic fragments
  51. - Fake rhetorical questions and colon reveals
  52. - Superficial analysis using "highlighting" or "underscoring"
  53. - Importance puffery and unsupported consensus claims
  54. - Generic filler, empty adverbs, and inflated buzzwords
  55. - Reaction narration after facts and statistics
  56. - Fake-profound and recap endings
  57. - Formatting habits such as decorative bold, unnecessary headings, emojis, and em dashes
  58. The full catalog lives in [`patterns.md`](skills/human-speak/references/patterns.md).
  59. The skill includes its own rules and final check. It does not need Nate's private workspace or a separate phrase list. Invoke it for an edit, audit, or draft; installing it does not apply an automatic filter to every response.
  60. ## Repository structure
  61. ```text
  62. .codex-plugin/plugin.json
  63. .claude-plugin/plugin.json
  64. .claude-plugin/marketplace.json
  65. skills/human-speak/SKILL.md
  66. skills/human-speak/references/patterns.md
  67. skills/human-speak/references/evaluation.md
  68. skills/human-speak/references/top-20.md
  69. ```
  70. ## Development
  71. Run the package checks:
  72. ```sh
  73. python scripts/validate_package.py
  74. ```
  75. For Claude Code, you can also run:
  76. ```sh
  77. claude plugin validate .
  78. claude --plugin-dir .
  79. ```
  80. ## Attribution
  81. Human Speak began as Nate Herk's personal AI Phrase Kill List and incorporates ideas from Peter Yang's MIT-licensed [No AI Slop](https://github.com/petergyang/no-ai-slop) skill. See [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md).
  82. ## License
  83. MIT

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.