Skip to content
Back to search

humanpen

SkillGitHubIncomplete

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/humanpen--humanpen-skill/badge.svg)](/listings/humanpen--humanpen-skill)

HTML

<a href="/listings/humanpen--humanpen-skill"><img src="/api/listings/humanpen--humanpen-skill/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.

<details open> <summary><b>Claude Code</b> (plugin)</summary>

/plugin marketplace add humanpen/humanpen-skill
/plugin install humanpen@humanpen

</details>

<details> <summary><b>Claude Code</b> (manual)</summary>

git clone https://github.com/humanpen/humanpen-skill ~/.claude/skills/humanpen

Project-scoped instead? Clone into `.claude/skills/humanpen` in the repo. </details>

<details> <summary><b>OpenAI Codex</b></summary>

git clone https://github.com/humanpen/humanpen-skill ~/.agents/skills/humanpen

Codex also reads `.agents/skills/` inside a repository, if you would rather scope it to one project. </details>

<details> <summary><b>CodeBuddy / WorkBuddy</b></summary>

git clone https://github.com/humanpen/humanpen-skill ~/.codebuddy/skills/humanpen

Project-scoped: clone into `.codebuddy/skills/humanpen` at the project root. </details>

<details> <summary><b>Cursor, Gemini CLI, OpenCode, Copilot, Goose, Amp, Kiro, and 35+ more</b></summary>

This is a plain [Agent Skill](https://agentskills.io): a folder with a `SKILL.md` in it. Every client that implements the standard loads the same folder — only the directory it scans differs, and each documents its own. Clone the repository into that directory and the skill is installed.

git clone https://github.com/humanpen/humanpen-skill

</details>

Prefer an MCP server? [humanpen-mcp](https://github.com/humanpen/humanpen-mcp) offers the same operations as tools, and works in Cursor, Windsurf, Cline, OpenCode and Claude Desktop as well.

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 · 15 lines
  1. {
  2. "name": "humanpen",
  3. "owner": {
  4. "name": "HumanPen",
  5. "email": "support@humanpen.net"
  6. },
  7. "plugins": [
  8. {
  9. "name": "humanpen",
  10. "description": "Lower AI-detection scores, fix citations, condense, and translate document files.",
  11. "version": "1.6.0",
  12. "source": "./"
  13. }
  14. ]
  15. }
README.md · README · 207 lines
  1. # humanpen-skill
  2. [![License](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](LICENSE)
  3. [![Agent Skill](https://img.shields.io/badge/Agent_Skills-open_standard-6E56CF.svg)](https://agentskills.io)
  4. [![Python](https://img.shields.io/badge/python-%3E%3D3.8-3776AB.svg)](https://python.org)
  5. English · [简体中文](README.zh-CN.md) · [日本語](README.ja.md)
  6. [Website](https://humanpen.net) · [Pricing](https://humanpen.net/pricing) · [Developer docs](https://humanpen.net/developers)
  7. > **Keywords:** ai humanizer, agent skill, turnitin ai detection, reduce ai score, humanize ai text, bypass ai detection, docx ai humanizer, ai content rewriter, ai writing tool, claude code skill, codex skill, ithenticate ai report
  8. **Humanize what's flagged. Preserve the rest.** An [Agent Skill](https://agentskills.io) for [HumanPen](https://humanpen.net) — a document-level AI humanizer that can humanize an entire document, rewrite user-selected passages, or automatically target flagged text from a Turnitin / iThenticate AI-detection report, editing `.docx` / `.pptx` files in place while preserving formatting, tables, images, citations, and formulas. Also converts citations between 12 styles, condenses to a word budget, and translates between 12 languages.
  9. ## Features
  10. - **Selective rewriting by detection report** — import a Turnitin / iThenticate AI Writing Report to pinpoint flagged passages; unflagged content is never touched
  11. - **Format in, format out** — a DOCX comes back as a DOCX, a PPTX as a PPTX; formatting, tables, images, and formulas survive intact and the result is still editable
  12. - **Academic structure preserved** — in-text citations, reference lists, footnotes, TOC fields, cross-references, figure numbering, equations, and special formatting are treated as protected objects
  13. - **No error injection** — restructures meaning and syntax to change expression; never adds grammar mistakes, spelling errors, or awkward sentences as a detection strategy
  14. - **Full-length documents** — no per-input word limit; a single file can be up to 100 MB, no splitting into text boxes
  15. - **Free to keep going** — still flagged? Re-humanize for free with a fresh report until the AI rate falls to `*` or 0%
  16. - **Word-count control (experimental)** — set a min/max word range to keep the output within a target length
  17. - **Pay per rewrite** — billed on words actually changed, not the whole document; failed and cancelled jobs cost nothing; credits never expire
  18. - **Nothing to install** — one Python script, standard library only, Python 3.8+
  19. ## Get a key
  20. Sign up at <https://humanpen.net> and create a key at
  21. <https://humanpen.net/settings/api-keys>. New accounts start with free credits,
  22. enough to put a document through and see what comes back.
  23. ```bash
  24. export HUMANPEN_API_KEY=hp_your_key
  25. ```
  26. Or copy `.env.example` to `.env` in this directory — it is gitignored.
  27. ## Install
  28. <details open>
  29. <summary><b>Claude Code</b> (plugin)</summary>
  30. ```
  31. /plugin marketplace add humanpen/humanpen-skill
  32. /plugin install humanpen@humanpen
  33. ```
  34. </details>
  35. <details>
  36. <summary><b>Claude Code</b> (manual)</summary>
  37. ```bash
  38. git clone https://github.com/humanpen/humanpen-skill ~/.claude/skills/humanpen
  39. ```
  40. Project-scoped instead? Clone into `.claude/skills/humanpen` in the repo.
  41. </details>
  42. <details>
  43. <summary><b>OpenAI Codex</b></summary>
  44. ```bash
  45. git clone https://github.com/humanpen/humanpen-skill ~/.agents/skills/humanpen
  46. ```
  47. Codex also reads `.agents/skills/` inside a repository, if you would rather
  48. scope it to one project.
  49. </details>
  50. <details>
  51. <summary><b>CodeBuddy / WorkBuddy</b></summary>
  52. ```bash
  53. git clone https://github.com/humanpen/humanpen-skill ~/.codebuddy/skills/humanpen
  54. ```
  55. Project-scoped: clone into `.codebuddy/skills/humanpen` at the project root.
  56. </details>
  57. <details>
  58. <summary><b>Cursor, Gemini CLI, OpenCode, Copilot, Goose, Amp, Kiro, and 35+ more</b></summary>
  59. This is a plain [Agent Skill](https://agentskills.io): a folder with a
  60. `SKILL.md` in it. Every client that implements the standard loads the same
  61. folder — only the directory it scans differs, and each documents its own. Clone
  62. the repository into that directory and the skill is installed.
  63. ```bash
  64. git clone https://github.com/humanpen/humanpen-skill
  65. ```
  66. </details>
  67. Prefer an MCP server? [humanpen-mcp](https://github.com/humanpen/humanpen-mcp)
  68. offers the same operations as tools, and works in Cursor, Windsurf, Cline,
  69. OpenCode and Claude Desktop as well.
  70. ## Commands
  71. ```bash
  72. # Lower AI-detection signals. --strategy: balanced | aggressive
  73. python3 scripts/humanpen.py humanize paper.docx --strategy balanced
  74. # Rewrite only the passages a detection report flagged
  75. python3 scripts/humanpen.py humanize paper.docx --report turnitin.pdf
  76. # Or only passages you choose: a JSON list of their exact texts
  77. python3 scripts/humanpen.py humanize paper.docx --segments-file passages.json
  78. # Constrain length (EXPERIMENTAL - a word limit weakens AI-rate reduction).
  79. # Whole document, or per-passage budgets for the report's flagged passages:
  80. python3 scripts/humanpen.py humanize paper.docx --min-words 1500 --max-words 1800
  81. python3 scripts/humanpen.py humanize paper.docx --report turnitin.pdf --segments-file budgets.json
  82. # Continue a finished job for FREE (once per job): re-run its still-flagged
  83. # passages with a fresh report for that job's result
  84. python3 scripts/humanpen.py free-rehumanize <job_id> --report new-turnitin.pdf
  85. # Citations and reference list to one style
  86. python3 scripts/humanpen.py fix-citations paper.docx --style apa7
  87. # Shorten to a word budget, or only chosen passages, each to its own budget
  88. # ([{"text": ..., "max_words": N}]). --mode delete drops whole low-value
  89. # sentences and keeps every other word verbatim (default: rewrite)
  90. python3 scripts/humanpen.py condense paper.docx --max-words 4000
  91. python3 scripts/humanpen.py condense paper.docx --segments-file budgets.json --mode delete
  92. # Translate, keeping layout
  93. python3 scripts/humanpen.py translate paper.docx --to zh
  94. # Read a Turnitin/iThenticate report: overall AI %, flagged passages
  95. python3 scripts/humanpen.py report turnitin.pdf
  96. # A job left running, and the balance
  97. python3 scripts/humanpen.py status <job_id> --download paper.docx
  98. python3 scripts/humanpen.py balance
  99. # Your jobs; cancel a queued or running one (its credits come back); run a
  100. # failed or cancelled one again (spends credits)
  101. python3 scripts/humanpen.py jobs --type humanize
  102. python3 scripts/humanpen.py cancel <job_id>
  103. python3 scripts/humanpen.py retry <job_id> --download paper.docx
  104. ```
  105. A finished job prints JSON with `output` (where the file landed) and
  106. `credits_charged` (what it cost). `--help` on any command lists its flags.
  107. Citation styles: APA 7, MLA 9, Harvard, Chicago (both), IEEE, Vancouver,
  108. GB/T 7714, AMA, ACS, OSCOLA. Translation covers 12 languages.
  109. ## Research projects
  110. Besides documents, the same key can hand a whole research, data analysis or writing task to HumanPen's cloud agents
  111. — the projects of the **Start a project** page: they scope the request with you, propose a plan, do the
  112. work, have it reviewed and deliver files. Create the key with **Research projects** ticked.
  113. ```bash
  114. # Start: upload the inputs, send the request; prints project_id and after_seq
  115. python3 scripts/humanpen.py project start "What drives churn? A 1,500-word report with the code." --file sales.csv
  116. # Wait until it needs you (questions, a plan to confirm) or is done
  117. python3 scripts/humanpen.py project wait <project_id> --after <after_seq>
  118. # Answer, confirm the plan you were shown, or take another allowed action
  119. python3 scripts/humanpen.py project act <project_id> confirm_plan --spec-version 1
  120. # Download the delivery as a zip; ask for changes with a message (a new round)
  121. python3 scripts/humanpen.py project download <project_id> -o out/
  122. python3 scripts/humanpen.py project message <project_id> "Add a section on pricing."
  123. ```
  124. Confirming a plan starts the paid work, charged by what it uses; there is no price up front. To let it run unattended, start it with a cap you are
  125. willing to spend: `--max-credits 20000 --auto-confirm --answer-recommended` confirms plans that fit and answers with
  126. the recommended answers, never past that cap for the whole project.
  127. ## Two things worth knowing
  128. **Jobs take minutes.** The command waits, printing progress to stderr.
  129. Interrupting it does not cancel the job — `status <job_id>` picks it back up.
  130. **`ai_percent` can be `null`, and that is usually good news.** Turnitin prints
  131. `*` instead of a number whenever AI writing comes in **under 20%** — it will not
  132. quantify that band, because too much of it is false positives. So `null` means
  133. "under 20%, and Turnitin will say no more", never "0%" and never "no result".
  134. ## Questions people ask
  135. **Will this bring a Turnitin AI score down?**
  136. Usually under 20% in one pass with `balanced` — the threshold below which
  137. Turnitin prints `*` instead of a number. If it misses, run the result back
  138. through with the new report; only the passages still flagged get rewritten.
  139. **Does it work with iThenticate too?**
  140. Yes — pass either report. The format is read from the file.
  141. **Is my document sent to the model?**
  142. No. It uploads the file and prints a path. A 40-page paper costs no tokens.
  143. ## Links
  144. - [API documentation](https://api.humanpen.net/v1/docs.md) ·
  145. [OpenAPI schema](https://api.humanpen.net/v1/openapi.json)
  146. - [humanpen-mcp](https://github.com/humanpen/humanpen-mcp) — the same
  147. operations as an MCP server
  148. - [humanpen.net](https://humanpen.net)
  149. Apache-2.0

Ratings and comments

Sign in to rate this skill.

Sign in

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