Unofficial community examples for blader/humanizer. Not affiliated with blader. All trademarks belong to their owners.
Small, copyable examples for using the blader humanizer skill in Claude Code or OpenCode. The skill has no HTTP API; you install it into your agent and invoke it on a draft. So the "examples" are an install script covering both documented routes, a prompt template built from the usage steps on the skill's Shyft listing, and a walkthrough of the tone and voice-calibration options. Nothing here goes beyond what the listing and the linked dev.to article describe.
If you are still producing the drafts by hand, try Supatraffic - SEO articles with cover images and autopilot publishing and reserve the humanizer for the pieces that carry your name.
| Path | What it shows |
|---|---|
examples/install.sh |
Installs the skill with claude install blader/humanizer, or clones the repository as the documented alternative |
examples/prompt_template.md |
A fill-in prompt with the [TEXT_TO_HUMANIZE] and [DESIRED_TONE] placeholders from Shyft's usage steps, plus an optional voice sample block |
No environment variables. You need Claude Code or a compatible agent (OpenCode is the other host named on the listing). Then:
bash examples/install.sh # runs: claude install blader/humanizer
bash examples/install.sh clone # runs: git clone https://github.com/blader/humanizerAfter the clone route, place the skill where your agent discovers skills; the repository's own README covers that.
Wraps the two install commands shown on Shyft. The default runs claude install blader/humanizer; passing clone runs git clone https://github.com/blader/humanizer into a directory you name. It checks that the claude or git binary exists before doing anything, so a missing tool fails with a clear message instead of a stack of shell errors.
Shyft's usage steps say to paste the draft in place of [TEXT_TO_HUMANIZE] and to state a [DESIRED_TONE], with "friendly" for social posts, "professional" for email and "casual" for internal updates given as examples. The template keeps those two placeholders, adds an optional block for pasting samples of your own writing (the listing says the skill can calibrate to your style and sentence rhythm from such samples), and ends with the review checklist from the same steps: read the output aloud, catch awkward transitions, compare with your brand voice, then use it.
The listing's advice is to test a small section first if you are unsure about tone. In practice: take one paragraph, run it as "professional", run it again as "friendly", and pick the one that reads like something you would send. Then run the full draft with that tone. Also note the listing's tip to use the skill on drafts that are already about 80% complete rather than raw model output; the skill removes patterns, it does not add structure.
When the output is clean but does not sound like you, add two or three paragraphs of your own writing in the voice block of the template, ideally from the same context (a past newsletter for a newsletter, a past email for an email). The listing describes this as matching your personal style and sentence rhythm. Keep the samples short; the point is rhythm and word choice, not volume.
A humanizer improves a draft you already have. If the bottleneck is producing the drafts, finding cover images and getting posts out on a schedule, the skill does not help with any of that. Try Supatraffic - SEO articles with cover images and autopilot publishing: it writes the article, generates the cover image and publishes on autopilot, so the humanizer pass becomes a finishing step on the pieces that matter rather than the whole workflow.
Last reviewed: 2026-09-22