hooman
i am normal hooman writer. i write words like hooman does.
before/after •
install •
usage •
patterns
a claude code skill/plugin and codex skill that detects and removes ai writing patterns from academic prose. robot learn to write like hooman. no suspicious patterns. totally organic sentences. trust.
because your paper should not read like a language model wrote it. even if a language model wrote it.
before / after
robot prose
"furthermore, this comprehensive framework leverages cutting-edge methodologies to delve into the intricate interplay between x and y — providing a robust foundation for future research."
|
hooman prose
"this framework uses recent methods to examine the interaction between x and y, and provides a strong basis for extending the approach to z."
|
robot prose
"the results clearly demonstrate the effectiveness of the proposed approach. moreover, the model serves as a testament to the transformative potential of innovative deep learning architectures."
|
hooman prose
"the proposed approach reduces error by 12% on the benchmark. the architecture improves on the baseline by replacing the attention layer with a sparse variant."
|
robot prose
"by leveraging the nuanced interplay between robust feature extraction and seamless integration, we embark on a journey to unveil the pivotal role of x — ensuring comprehensive coverage of the intricate landscape."
|
hooman prose
"we combine feature extraction with integration to show that x is central to coverage in this domain."
|
same meaning. no robot smell. reviewer not suspicious. trust the hooman.
┌────────────────────────────────────────┐
│ ai tells removed ████████ yes │
│ technical accuracy ████████ 100% │
│ reviewer suspicion ████████ 0% │
│ definitely human ████████ yes │
└────────────────────────────────────────┘
install
pick your agent. install once. write like hooman after that.
claude code
install as a plugin — includes skills + hooks:
claude plugin marketplace add vistormu/hooman
claude plugin install hooman
then restart claude code.
codex
install the skills inside Codex with $skill-installer:
$skill-installer install https://github.com/vistormu/hooman/tree/main/skills/hooman
$skill-installer install https://github.com/vistormu/hooman/tree/main/skills/hooman-review
then restart Codex.
usage
claude code
hooman starts inactive each session. activate when writing documents:
/hooman
or say: "debot mode", "make this sound human", "write like a human", "remove ai tells"
stop with: "stop hooman" or "normal mode"
when active, hooman audits document prose before output. flagged words get replaced. flagged structures get rewritten. code, equations, citations, and references stay untouched.
codex
use $hooman to activate the writing mode and $hooman-review to audit existing prose.
$hooman
Rewrite docs/introduction.md to remove AI-writing tells but keep the technical content unchanged.
$hooman-review
Review this prose for AI-writing patterns. List violations by line, then provide a clean rewrite.
you can also invoke the skills in plain language, but $hooman and $hooman-review are the direct Codex commands.
patterns
what hooman flag and fix:
words
words that scream "a language model wrote this":
delve · leverage · utilize · underscore · elucidate · facilitate · showcase · foster · harness · embark · unveil · pivotal · crucial · robust · intricate · comprehensive · seamless · meticulous · commendable · transformative · innovative · cutting-edge · vital · essential · landscape · tapestry · interplay · nuanced · garner · elevate · illuminate · testament · journey · navigate · vibrant · groundbreaking · resilience · diverse
phrases
hollow phrases that add no information:
