codex-mcp-server-integration
Warn
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs users to install external packages
@openai/codexandcodex-mcp-serverfrom the npm registry. These packages originate from unverified sources and are not associated with a trusted vendor, posing a potential supply-chain risk. - [CREDENTIALS_UNSAFE]: Installation instructions advise users to authenticate the Codex CLI by providing their
OPENAI_API_KEY. Supplying sensitive credentials to an unverified third-party tool whose security posture has not been established is a significant risk. - [COMMAND_EXECUTION]: The skill functions as a wrapper for a local CLI that supports autonomous mode (
fullAuto: true) and file system write access (sandbox: "workspace-write"). This capability allows remote instructions to potentially perform unauthorized modifications to the user's workspace. - [DATA_EXFILTRATION]: Tools within the skill, such as
codexandreview, ingest local source code files and pass them to an external CLI. This pattern facilitates the transmission of potentially proprietary or sensitive code to external services and unknown endpoints.
Audit Metadata