You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
checked it. the "php backdoor" is just example webshell code inside the SKILL.md docs (offensive-file-upload / offensive-persistence) — it's documentation of the technique, inert text in markdown, not a real backdoor. installing the skills doesn't give anyone remote access to you. install.sh and the python tooling are clean too — no curl|bash, no network calls, no persistence. that webshell string is what trips AV scanners (Defender flags it as Perhetshell), but it's a signature match on tutorial code, not an actual planted backdoor
the one thing actually worth a look is the skill-optimize-apply.yml workflow — comment-triggered with contents: write. that's a repo-side CI risk, not a user-facing backdoor, but different topic than the issue.
Activity
Adamvann commentedon Sep 23, 2026
checked it. the "php backdoor" is just example webshell code inside the SKILL.md docs (offensive-file-upload / offensive-persistence) — it's documentation of the technique, inert text in markdown, not a real backdoor. installing the skills doesn't give anyone remote access to you. install.sh and the python tooling are clean too — no curl|bash, no network calls, no persistence. that webshell string is what trips AV scanners (Defender flags it as Perhetshell), but it's a signature match on tutorial code, not an actual planted backdoor
the one thing actually worth a look is the skill-optimize-apply.yml workflow — comment-triggered with contents: write. that's a repo-side CI risk, not a user-facing backdoor, but different topic than the issue.