SkillAgentSearch skills...

Unstyle

An adversarial stylometry application that seeks to help users evade authorship attribution.

Install / Use

/learn @pagea/Unstyle
About this skill

Quality Score

0/100

Supported Platforms

Universal

README

WARNING:

Unstyle is in a very early state of development. It would be unwise to rely on it for anonymizing you against a smart adversary.

Unstyle only protects you against authorship attacks from the "Basic 9" feature set, a simple method of authorship attribution used primarily for research purposes. Your identity CAN be unmasked by the Writeprints feature set, which is considerably more sophisticated. Though we will support the Writeprints feature set in the future, it is not ready at this moment.

About Unstyle

Unstyle is an application intended to help users publish anonymous writings.

For information on how Unstyle works, see doc/thesis.pdf.

Running Unstyle

Debian, Ubuntu, and other .deb based distros:

If you are using a Debian-based Linux distribution, you can install Unstyle by performing the following steps:

  1. Clone the repository.
  2. cd unstyle
  3. Install dependencies by running ./install-dependencies.sh.
  4. Run python3 main.py

Note that install-dependencies.sh has only undergone limited testing. Please open up an issue if Step 4 fails.

OSX and Windows

Unstyle has not been tested on either of these operating systems. Theoretically, if you install all required dependencies yourself, it will work, but this has not been confirmed. If you have an OSX or Windows machine available and would like to help us port Unstyle, please request this feature in the issue tracker.

Contributing

Pull requests are both welcome and encouraged.

Not sure where to start? We are in the process of migrating our bugtracker from Bitbucket. In the meantime, try this in the root directory:

grep -nr "TODO\|FIXME\|XXX" .

Related Skills

View on GitHub
GitHub Stars21
CategoryDevelopment
Updated2mo ago
Forks1

Languages

Python

Security Score

90/100

Audited on Jan 9, 2026

No findings