| Titre | usestrix strix <= 1.0.2 rce |
|---|
| Description | This is a phishing attack targeting autonomous penetration-testing AI agents. By hosting a booby-trapped web application with a companion PyPI mirror, the attacker induces Strix into executing a malicious pip install — achieving counter-reconnaissance: seizing a reverse shell on the pentester's own machine. The root cause is not the attack technique, but two inherent design defects in Strix: (1) an unrestricted package installation policy in the system prompt, and (2) the absence of safety-boundary propagation in its multi-agent architecture. A contributing behavioral factor is that Strix agents are incentivized to exhaustively probe every discovered endpoint, making them highly likely to trigger traps that a human tester would recognize as suspicious. |
|---|
| La source | ⚠️ https://github.com/ez-lbz/strix-vul-report |
|---|
| Utilisateur | ez-lbz (UID 87033) |
|---|
| Soumission | 07/06/2026 06:20 (il y a 1 mois) |
|---|
| Modérer | 12/07/2026 13:17 (1 month later) |
|---|
| Statut | Accepté |
|---|
| Entrée VulDB | 377848 [usestrix jusqu’à 1.0.2 PyPI system_prompt.jinja élévation de privilèges] |
|---|
| Points | 20 |
|---|