OEBP defines semantic behavior contracts for embodied systems. Invalid, untrusted, or malicious behavior documents can create physical risk when they are connected to real robots. Treat every external behavior proposal, adapter, capability profile, context source, and registry entry as untrusted until it passes deterministic validation and the deployment's safety supervisor.
Please open a private security advisory or contact the maintainers before publicly disclosing vulnerabilities that could enable unsafe robot behavior, validator bypass, registry poisoning, adapter abuse, or unsafe generated data.
If private advisories are unavailable for the repository, open a minimal public issue that says a security report exists without publishing exploit details.
- Validator bypasses.
- Ambiguous semantics that can cause unsafe execution.
- Adapter or registry spoofing.
- Capability-profile misrepresentation.
- Unsafe handling of model-generated behavior proposals.
- Trace or provenance tampering.
- Supply-chain risks in reference implementations.
- Unsupported forks or private downstream deployments.
- Vulnerabilities in unrelated robot middleware unless OEBP integration makes the issue materially worse.
- Physical safety guarantees for hardware that does not implement the required external safety supervisor.
OEBP validation is necessary but not sufficient for real-world execution. A deployment MUST keep independent safety controls, emergency stop behavior, hardware limits, operator policy, and environment-specific risk checks outside the protocol runtime.