Agent2Canoe currently provides security fixes for the latest 0.3.x release line. Older development previews and the 0.1.x/0.2.x lines should be upgraded before reporting a version-specific defect.

Do not open a public issue containing credentials, API keys, client certificates, Seed/Key material, proprietary CANoe configuration data, or a working exploit. Use the repository's private GitHub vulnerability reporting channel when available. If private reporting is unavailable, contact the repository owner through their GitHub profile and request a private channel without including sensitive details in the initial message.

Please include:

• affected Agent2Canoe version and commit;
• Windows, Python, CANoe and transport details;
• affected interface: Python, REST, or MCP;
• required role and authentication method;
• minimal reproduction steps and observed impact;
• whether the issue crosses confirmation, identity, lease, redaction, path, or server-owned Provider boundaries;
• sanitized logs with tokens, keys, environment values, paths, vehicle data, and proprietary identifiers removed.

The maintainer will acknowledge a complete report, reproduce it privately, agree on disclosure timing, and publish a fixed release and advisory when the impact is confirmed. Do not test against systems, CANoe projects, ECUs, or hardware you do not own or have explicit permission to assess.

• AI clients never receive direct CANoe COM authority from the automation server and cannot self-approve operations.
• Side-effecting calls require role checks and parameter-bound, short-lived, single-use confirmation tokens.
• UDS SecurityAccess algorithms and secrets remain in server-owned Providers; generic diagnostic service 0x27 bypass attempts are rejected.
• Export paths are server-owned and bounded; clients submit file names, not arbitrary filesystem destinations.
• Physical hardware-channel writes are not exposed in v0.3.0.

See identity and mTLS, confirmation tokens, and SecurityAccess providers for deployment details.