Lucene search
+L

5 matches found

osv_cve
osv_cve
•added 2026/10/20 12:46 a.m.•18 views

CVE-2026-77930 EvoScientist 0.1.0 before 0.2.5 Human Approval Bypass via Async Sub-Agents

EvoScientist 0.1.0 before 0.2.5 contains an approval bypass vulnerability that allows attackers to run host shell commands by delegating them to async sub-agents lacking an interrupton gate. Attackers can plant instructions in ingested content such as fetched web pages, so the agent calls...

9.3CVSS6AI score
SaveExploits0References10
Vulnrichment
Vulnrichment
•added 2026/10/20 12:46 a.m.•10 views

CVE-2026-77930 EvoScientist 0.1.0 before 0.2.5 Human Approval Bypass via Async Sub-Agents

EvoScientist 0.1.0 before 0.2.5 contains an approval bypass vulnerability that allows attackers to run host shell commands by delegating them to async sub-agents lacking an interrupton gate. Attackers can plant instructions in ingested content such as fetched web pages, so the agent calls...

9.8CVSS5.9AI score0.00688EPSS
SaveExploits0References6
Cvelist
Cvelist
•added 2026/10/20 12:46 a.m.•25 views

CVE-2026-77930 EvoScientist 0.1.0 before 0.2.5 Human Approval Bypass via Async Sub-Agents

EvoScientist 0.1.0 before 0.2.5 contains an approval bypass vulnerability that allows attackers to run host shell commands by delegating them to async sub-agents lacking an interrupton gate. Attackers can plant instructions in ingested content such as fetched web pages, so the agent calls...

9.8CVSS0.00688EPSS
SaveExploits0References6
CVE
CVE
•added 2026/10/20 12:46 a.m.•32 views

CVE-2026-77930

EvoScientist versions 0.1.0 before 0.2.5 are vulnerable to an approval bypass vulnerability. The root cause is the use of async sub-agents that lack an interrupt_on gate, which allows attackers to execute arbitrary host shell commands. By planting malicious instructions in ingested content, such ...

9.8CVSS5.9AI score0.00688EPSS
SaveExploits0References6
EUVD
EUVD
•added 2026/09/21 8:45 p.m.•31 views

EUVD-2026-84161

MaxKB is an open-source AI assistant for enterprise. Prior to version 2.10.5-lts, assistants with a tool, MCP tool, skill, or sub-application use SandboxShellBackend, which exposes an execute shell tool without excluding it and omits execute from interrupton, so human approval is not required...

10CVSS6.3AI score0.01049EPSS
SaveExploits1References3
Rows per page
Query Builder