# Humanizer: This One Is Plainly Not Just for Developers

Date: 2026-10-21

This is the post that weakens the series thesis. Humanizer is a writing-quality skill. Its primary users are not engineers reviewing pull requests. The honest response to that is to say so, and then explain what it actually offers.

<!--more-->

## What the skill does

Humanizer (version 3.1.0) rewrites AI-sounding text so it reads like the writer wrote it, without changing what it says. It is built on Wikipedia's "Signs of AI writing," the guide Wikipedia editors use to catch AI-generated text, and it installs into Claude Code, Codex, Claude.ai, Claude Desktop, and any agent the Skills CLI supports.

The skill documents 26 patterns that mark AI-generated prose. The five strongest—each justifying an edit on a single sighting—are:

1. **Not X but Y.** "It's not just a feature, it's a shift." The negative half names something no one claimed, so the positive half sounds larger than it is.
2. **One-line closers.** "Let that sink in." "That is the real win." A sentence that asks the reader to pause on a claim instead of adding to one.
3. **Sayings that sound deep.** "At its core, what really matters is..." An ordinary point dressed as a hidden truth.
4. **Staged run-up.** "Here's the thing." "Honestly?" An announcement of the point in place of the point.
5. **Arguing with no one.** "I'm not saying X, but..." An objection the reader never raised.

The remaining 21 patterns cover rhythm by rule, inflation, formatting by rule, leftover chat residue ("Great question!"), and re-explaining what the reader already knows. Each is documented with a before/after pair.

## Three patterns that appear in real product communications

**Inflated announcements.** The README's documented example: "I'm thrilled to announce that shared drafts are finally here! 🚀" becomes "Shared drafts are out today." The rewrite contains the same information. The original adds enthusiasm and removes one word of actual content ("today").

**Generic dramatic closers.** Pattern 2: a one-sentence paragraph that restates the preceding section. "That is the real win." appearing after three consecutive sections is the same tell at paragraph scale—the same signal applied by rule rather than by meaning.

## An illustrative example

**Before:**
> We are absolutely thrilled to share that your reporting dashboard has just gotten a major upgrade. This is a pivotal moment for our team, and more importantly, for yours. You can now filter by date range, team member, and project type—all in one place. That's not just a feature; it's the visibility you've been asking for. That is the real win.

**After:**
> Your reporting dashboard now lets you filter by date range, team member, and project type at the same time. The update is live today.

Four patterns removed—inflated announcement, inflated significance, a Not-X-but-Y contrast, and a one-line closer—and two facts remain. No product benefit was added in the rewrite.

## Who benefits

Founders writing product announcements, communications teams drafting external releases, and technical writers editing documentation. Developers producing user-facing content—release notes, changelogs, onboarding guides—are also a natural fit. The skill accepts a writing sample to match voice: rhythm, word choice, punctuation, and deliberate quirks.

## Limitations

Cleaner prose does not establish accuracy. If the underlying claim is wrong, a well-edited sentence is a well-edited wrong claim. The SKILL.md states: "Do not add a fact, name, number, date, quote, or citation unless it comes from the source or the user."

On AI detectors: the README states directly—"Getting past AI detectors is not a goal, and detectors still flag most of its output." The repository is not permission to misrepresent authorship.

## Where this sits in the series

Using an agent-based workflow to edit prose does not make the user's work development work. Humanizer runs in a coding agent, but the task is writing. For this tool, the relevant question is not whether the user is a developer but whether the agent-based workflow offers enough value over a simpler editing approach—a single well-crafted prompt, a style guide, or a direct revision of the text—to justify the setup.

For high-volume outputs that consistently show the same patterns, the case is cleaner. For occasional editing, a human reviewer who knows the 26 patterns can apply the same logic without an agent.

## Next

Humanizer addresses individual documents. The next tool in this series, Graphify, addresses a different problem: what happens when the knowledge a team needs is scattered across dozens of files, and no single document contains the answer?

---

*Source: github.com/blader/humanizer — README and SKILL.md (version 3.1.0). Retrieved 8 October 2026.*

#Writing #ContentStrategy #AgentSkills