CVE-2026-19060
MediumCVSS 5.3Exploitation Probability (EPSS)
Low risk9th percentile - higher than 9% of all known CVEs
Summary
A vulnerability was identified in FoundationAgents MetaGPT up to 0.8.2 affecting an unknown function, allowing code injection. The attack requires local access. The exploit is publicly available and might be used. The vendor did not respond.
Risk Assessment
An attacker with local access can execute arbitrary code, potentially leading to system compromise or data theft.
Recommendation
It is recommended to restrict local access, monitor activities, and apply any available patches or workarounds if they exist.
Other vulnerabilities in MetaGPT
See all- CVE-2026-19059Low
A path traversal vulnerability was found in FoundationAgents MetaGPT up to version 0.8.2, affecting the read function in metagpt/tools/libs/editor.py. The attack requires local access. The exploit has been publicly disclosed and may be used. The vendor did not respond to the disclosure.
- CVE-2026-19058Medium
A vulnerability was found in FoundationAgents MetaGPT up to 0.8.2 in the DataInterpreter function of the file metagpt/roles/di/data_interpreter.py, allowing code injection. The attack must be initiated locally. The exploit is public and could be used. The vendor did not respond.
- CVE-2026-11455Medium
A vulnerability has been determined in FoundationAgents MetaGPT up to version 0.8.2, allowing command injection in the check_cmd_exists function of metagpt/utils/common.py. The attack may be initiated remotely but requires high complexity. The exploit is publicly available, and the project has not responded.
- CVE-2026-10566Medium
In MetaGPT up to version 0.8.2, the function Message.check_instruct_content in metagpt/schema.py is vulnerable to deserialization via manipulation of the mapping argument. The attack requires local access.
Original NVD description (English source)
A vulnerability was identified in FoundationAgents MetaGPT up to 0.8.2. This impacts an unknown function. Such manipulation leads to code injection. The attack needs to be performed locally. The exploit is publicly available and might be used. The vendor was contacted early about this disclosure but did not respond in any way.