CVE-2026-77930
8.8HighPublished 20 Oct 2026 · NVD status: Received · AI triage by gemini-3.8-flash
- Attack vector
- Network
- Complexity
- Low
- Privileges needed
- None
- User interaction
- Required
- Impact
- Full C/I/A compromise
- Weakness
- CWE-862
Impact
Attackers can execute arbitrary shell commands with user privileges and steal environment secrets. To trigger this, an attacker must plant malicious instructions in ingested content, such as fetched web pages, that the agent processes.
Root cause
Async sub-agents spawned via start_async_task fail to enforce an interrupt_on gate, bypassing required human approval checks.
Affected systems
EvoScientist 0.1.0 before 0.2.5
CVSS v3.1 · CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Fix & remediation
Upgrade EvoScientist to version 0.2.5 or later. If upgrading immediately is not possible, stop the agent from fetching untrusted web pages or parsing external content.
$ pip install --upgrade evoscientist>=0.2.5AI-suggested. Run it in staging first and check it against the vendor advisory. Why?
NVD description
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 interrupt_on gate. Attackers can plant instructions in ingested content such as fetched web pages, so the agent calls start_async_task and runs commands with user privileges and environment secrets.
References
- https://github.com/EvoScientist/EvoScientist/blob/ac58caab7bdd1706f3fee55891f611cd175d05b0/EvoScientist/subagents/_factory.py#L116
- https://github.com/EvoScientist/EvoScientist/commit/d249e320bd856a7d31d859cfbb2bf75bab6c8686
- https://github.com/EvoScientist/EvoScientist/issues/387
- https://github.com/EvoScientist/EvoScientist/releases/tag/v0.2.5
- https://www.lasso.security/vulnerability-database/cve-2026-77930
- https://www.vulncheck.com/advisories/evoscientist-0.1.0-before-0.2.5-human-approval-bypass-via-async-sub-agents