In nanobot, when sandboxed exec is enabled, the shell executable used to run commands can be selected by the caller and is launched before the sandbox is applied. An attacker who can induce the agent to create an executable in the workspace and select it as the shell can cause that executable to run outside the sandbox (e.g. Linux bubblewrap), with the operating-system permissions of the nanobot process. This issue affects nanobot (nanobot-ai) from 0.2.1 before 0.4.1.
Casky was already ahead
This CVE exploits attack patterns that Casky's 0matched skills already investigate — long before this vulnerability was disclosed. Claude's reasoning model maps these techniques to MITRE ATT&CK, so practitioners who ran these skills have already seen the threat behaviour in their findings.
CVE-2026-121864 represents a critical sandbox escape vulnerability in nanobot versions 0.2.1 through 0.4.0, where the sandboxed execution feature fails to properly enforce security boundaries. The vulnerability allows an attacker to specify an arbitrary shell executable that runs before sandbox restrictions (such as Linux bubblewrap) are applied, enabling arbitrary code execution with the full OS-level permissions of the nanobot process. This affects any organization deploying nanobot as an AI agent framework for automated task execution, particularly in multi-tenant environments or where untrusted input can influence command execution paths. The high CVSS score of 8.5 reflects the severity of privilege escalation and sandbox bypass capabilities an attacker gains.
While MITRE ATT&CK technique mappings are not formally assigned to this CVE, the attack pattern aligns with Privilege Escalation and Defense Evasion tactics—specifically execution flow manipulation before security controls activate. Casky's 754 security skills, powered by Claude AI's extended reasoning capabilities, would detect this vulnerability through behavioral analysis of: (1) executable creation patterns in workspace directories followed by shell selection calls, (2) timing discrepancies between sandbox initialization and process launch, and (3) permission escalation anomalies where child processes inherit elevated privileges despite sandboxing claims. Practitioners monitoring nanobot deployments would observe findings flagging suspicious shell executable selection from user-controlled paths, pre-sandbox process spawning attempts, and capability leakage indicators—allowing security teams to identify exploitation attempts before sandbox escape occurs.
Composite risk scoring from EPSS, CISA KEV, Shodan, and GreyNoise — 21 security APIs correlated into a single Casky Risk Score. Coming in Casky Pro. Join early access →
Casky has 0 skills that investigate the attack patterns behind CVE-2026-121864. Run one and get CVSS-scored findings in 3 minutes.
Run the skill that detects this →© 2026 Casky.AI, Inc. · AI Security Investigation