tech-humanizer-skill
High-signal technical writing. No AI fluff.
Stop shipping PRs and docs that sound like they were written by a generic assistant. tech-humanizer-skill is an agent-side tool that removes AI-writing markers (ceremonial framing, vague claims, and promotional filler) while protecting the technical precision engineers rely on. It targets the right output format whether that's a Slack update or a formal engineering design doc.
Why this matters
- Protects technical depth: shields terms like
Kubernetes,OAuth 2.0,JWT, andCI/CDfrom being simplified into vague filler. - Grounded in methodology: uses a community-driven field guide based on Wikipedia's AI markers to identify and fix specific tells.
- Learns your voice: maintains a local
writing-profile.jsonthat tracks your team's nomenclature and personal style preferences. No black boxes. - Context-aware: a PR description needs impact; a Slack message needs brevity; a technical doc needs precision.
Before and after: real engineering scenarios
| Context | Generic AI output | Humanized output |
|---|---|---|
| Engineering (PR) | "This pull request encompasses a comprehensive refactoring of the authentication module to enhance security posture." | "Refactored the auth module to fix a potential session hijacking vulnerability." |
| Product (Release Notes) | "We are excited to unveil a plethora of new features designed to empower your workflow and drive synergy." | "We've added bulk-editing and a new dashboard to speed up your daily tasks." |
| Architecture doc | "It is important to note that our innovative solution leverages Kubernetes to optimize scalability and ensure high availability." | "We use Kubernetes to handle scaling and keep the service up during traffic spikes." |
| On-call / chat | "I am reaching out to inform you that I have completed the requested investigation into the latency issues." | "Found the cause of the latency spikes. It was a missing index on the users table." |
| Email to client | "Please find the attached documentation for your perusal. Should you have any questions, feel free to reach out." | "I've attached the docs. Let me know if you have questions." |
Who this is for
- Individual contributors keeping PRs high-signal and author-like.
- Leads and architects keeping technical docs precise and free of assistant-style filler.
- Product owners writing release notes that tell users what changed.
- Teams standardizing internal nomenclature via a shared writing profile.
How it works
- Diagnostic scan: identifies markers like ceremonial framing (e.g., "It is important to note") and vague attribution (e.g., "Many believe").
- Technical shielding: cross-references with local lexicons to keep protected terms verbatim.
- Recursive check: re-evaluates the output against a rubric to confirm the result reads like a human wrote it.
Installation
Add to your project
npx skills add clydeshen/tech-humanizer-skill
Update to the latest version
npx skills update clydeshen/tech-humanizer-skill
Supported agent tools
This skill uses the standard SKILL.md layout and is compatible with:
- Claude Code (native support)
- Gemini CLI (built-in integration)
- Codex and Kiro
- Claude Desktop
Writing profile: local and transparent
Unlike cloud humanizers, this skill uses a local writing-profile.json in your project root. It is private, ignored by Git, and lets you:
- Define nomenclature: "Always use Registry, never Store."
- Store samples: give the skill a few of your own sentences to match your register.
- Internalize corrections: if you correct an AI word twice, the skill will not use it again.
Project links
- Contributing: help expand the AI-marker field guide.
- Code of Conduct: standard community guidelines.
- Security: how we handle vulnerability reports.
- MIT License: open-source and free.
No comments yet
Be the first to share your take.