tech-humanizer-skill

skills.sh skills.sh License PRs Welcome Compatible with Claude Code Compatible with Gemini CLI

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, and CI/CD from 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.json that 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

  1. Diagnostic scan: identifies markers like ceremonial framing (e.g., "It is important to note") and vague attribution (e.g., "Many believe").
  2. Technical shielding: cross-references with local lexicons to keep protected terms verbatim.
  3. 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