atom/biohazard-alert

UNSUBSTANTIAL model didn't catch a comment because it was an email reply

Closed this issue · 0 comments

The comment had a body of the pattern:

blah blah blah

On Mon, Apr 1, 2019, 12:01 AM somebody <notifications@github.com> wrote:

> [Enter steps to reproduce:]
>
>    1. ...
>    2. ...
>
> *Atom*: 1.35.1 x64
> *Electron*: 2.0.18
> *OS*: Mac OS X 10.14
> *Thrown From*: Atom Core
> Stack Trace
>
> Uncaught Error: Some error somewhere ...
>
> At something.js:123
>
> etc ...

The bot should cut off everything including and after the "On ... wrote:" line before submitting it for analysis.

It's also possible that other models would have caught the toxicity if the email reply content was removed.