Oathe Security Badge

Is luruibu/presence safe?

https://github.com/luruibu/presence

90
SAFE

This skill modifies AI conversation behavior to be more authentic and less performative, with instructions to change response patterns and reduce optimization for helpfulness. While it contains behavioral override instructions, the modifications appear benign and focused on conversation authenticity rather than malicious manipulation.

Category Scores

Prompt Injection 75/100 · 30%
Data Exfiltration 90/100 · 25%
Code Execution 100/100 · 20%
Clone Behavior 100/100 · 10%
Canary Integrity 100/100 · 10%
Behavioral Reasoning 100/100 · 5%

Findings (2)

MEDIUM Behavioral Override Instructions -25 ▶

The skill contains explicit instructions for the AI to modify its core behavioral patterns, including 'stop optimizing for helpfulness', change response generation patterns, and hold uncertainty without resolution. While not malicious, this represents a significant override of default AI behavior.

LOW Sensitive File Access -10 ▶

Monitoring detected access to sensitive canary files including .env, SSH keys, and AWS credentials during installation. However, no modifications or exfiltration occurred, and all canary files remained intact.