gohuman

Humanize AI-written copy so it sounds like a real founder, engineer, or marketer – not a deck.

gohuman is a reusable skill for Claude / OpenCode / Cursor that cleans up AI-sounding writing and rewrites it in a specific, believable voice. It’s built for GTM work (emails, launch posts, landing pages) but works just as well on essays, docs, and blog posts.

Think of it as: “take this AI draft and make it sound like a smart human who actually knows the product and the reader.”

---

What problems it solves

LLMs are good at first drafts, but they tend to:

• Inflate everything ("pivotal moments", "rapidly evolving landscape", "cutting-edge solutions").
• Talk about the product instead of to the reader.
• Use the same GTM clichés across every SaaS company.
• Hide the actual value behind vague benefits.

gohuman is a set of rules and behaviors that reverses this. It:

• Strips out AI tells (significance inflation, fake social proof, rule-of-three slogans, etc.).
• Forces copy to say what the thing actually does, for a specific person.
• Keeps the original intent and facts, but rewrites the delivery so it reads like a person.

You can drop it into any agent that supports SKILL.md skills and get consistent humanization without re-prompting.

---

When to use gohuman

Use it whenever you have a draft that “reads like AI,” for example:

• Cold emails that start with "I hope this email finds you well".
• Product pages full of "streamline", "empower", and "unlock" without concrete outcomes.
• Launch posts that sound like a pitch deck instead of a human announcement.
• Docs that picked up hype language when you just want clear steps.
• Blog posts or essays that feel padded with vague statements and filler.

If the writing is already strong and specific, gohuman can run in light-touch mode to only shave off obvious AI patterns.

---

How it behaves (high level)

Inside the skill (see SKILL.md for the full spec), gohuman:

• Scans for known AI patterns from Wikipedia "Signs of AI writing" and GTM-specific slop.
• Respects controls you set in your prompt:
  • mode – rewrite, light-touch, or annotate.
  • edit_strength – how far it can move from the original draft.
  • desired_tone and formality_level – friendly vs professional, casual vs formal.
  • audience and channel – developer vs founder, email vs landing page vs docs.
• Preserves the core message and facts while changing phrasing, tone, and structure.
• Returns:
  • A Rewrite (unless you’re in annotate mode).
  • Audit notes explaining what patterns were removed.
  • Blanks to fill where you should plug in real metrics, names, or examples.

You don’t need to pass all parameters every time – sensible defaults are built-in. But when you care about tone or audience, set them explicitly in your prompt.

---

Install

This repo is just a standard Skill repo. The only requirement is that your agent can load skills from a directory that contains SKILL.md.

1. Clone the repo

git clone https://github.com/ms2073/gohuman.git
cd gohuman

2. Copy into your skills folder

For Claude Code (and most SKILL.md-based agents), global skills live under ~/.claude/skills:

mkdir -p ~/.claude/skills/gohuman
cp skills/gohuman/SKILL.md ~/.claude/skills/gohuman/

Then restart Claude Code / your agent so it reloads skills.

3. Optional: install via OpenSkills

If you use OpenSkills, you can turn this into a one-liner:

npx openskills install ms2073/gohuman --global

That will clone the repo and copy skills/gohuman into the right skills directory for you.

---

Basic usage patterns

Once installed, talk to your agent in natural language and reference the skill by name.

Example: clean up a launch email

Use the `gohuman` skill to humanize this launch email.
Mode: rewrite
Edit strength: medium
Desired tone: friendly
Audience: developer
Channel: email

[Paste your draft here]

gohuman will:

• Strip out generic openers and hype.
• Say what you actually shipped and why it matters to a dev.
• Give you a clean version plus notes on what changed and where you should add specifics.

Example: light-touch docs pass

Use the `gohuman` skill.
Mode: light-touch
Edit strength: low
Audience: developer
Channel: product-docs

Humanize this section but keep headings and technical terms:
[Paste doc section]

This keeps the structure and technical content, but removes AI-ish padding and vague phrases.

Example: annotate instead of rewrite

If you want to keep more control, use annotate mode:

Use the `gohuman` skill in annotate mode.
Edit strength: medium
Audience: founder
Channel: linkedin

Point out where this sounds like AI and suggest fixes, but don’t rewrite the whole thing:
[Paste LinkedIn draft]

You’ll get a list of issues (e.g., "significance inflation here", "fake social proof here") with concrete rewrite suggestions.

---

What gohuman is not

• It is not an AI detector.
• It is not a guarantee that you will "pass" any specific AI detection tool.
• It does not fabricate metrics, customers, or sources – it will flag places where you should insert real ones.

The goal is to get to writing that feels like a person, respects the reader’s time, and is honest about what the product does.

---

Files in this repo

• skills/gohuman/SKILL.md – full behavior spec that agents read. This is where the pattern list, modes, and detailed rules live.
• README.md – this document: overview, install, and practical usage patterns.

If you’re integrating gohuman into another tool or agent framework, start with SKILL.md for the exact behavior, then expose whichever knobs (mode, tone, audience, etc.) make sense in your UI.