OpenHands — agentic threat model
OpenHands possesses a high-risk agentic profile due to its ability to execute arbitrary shell commands and modify codebases autonomously. Without verified sandboxing and strict human-in-the-loop controls, compromise could lead to severe host exploitation or supply chain attacks.
OWASP AIVSS score rationale
| Autonomy of Action | 0.80 | |
| Goal-Driven Planning | 0.80 | |
| Self-Modification | 0.30 | |
| Dynamic Tool Use | 0.90 | |
| Persistent Memory | 0.50 | |
| Contextual Awareness | 0.80 | |
| Dynamic Identity | 0.20 | |
| Multi-Agent Interactions | 0.30 | |
| Non-Determinism | 0.70 | |
| Opacity & Reflexivity | 0.60 |
Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.
MAESTRO 7-layer threat model
Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.
Not certain from the listing — The listing does not specify the foundation models used, but as an open-source coding assistant, it likely supports multiple LLMs (e.g., GPT-4, Claude, or local models), making it susceptible to prompt injection, adversarial reprogramming, or misaligned code generation if the underlying model is compromised.
Not certain from the listing — The listing does not detail data operations or vector stores. However, a backlog-tackling agent typically requires access to codebase repositories, raising risks of data exfiltration or codebase poisoning if malicious code is introduced into the training/RAG pipeline.
As an autonomous coding assistant, OpenHands orchestrates complex planning and tool execution (e.g., running bash commands, editing files). This introduces severe risks of tool misuse, where prompt injection could lead to arbitrary code execution or unauthorized file modification within the workspace.
Not certain from the listing — The listing does not specify the deployment infrastructure or sandboxing mechanisms (such as Docker containers). Without robust sandboxing, executing agent-generated code or bash commands poses extreme risks of host compromise, privilege escalation, and lateral network movement.
Not certain from the listing — There is no mention of built-in guardrails, evaluation frameworks, or real-time observability tools. A lack of monitoring could allow malicious or buggy agent actions to go unnoticed until code is committed or executed.
Not certain from the listing — The listing does not outline access controls, authentication, or compliance frameworks. Operating directly on developer backlogs requires strong repository access controls and audit logging to prevent unauthorized code commits.
Not certain from the listing — The listing does not describe multi-agent coordination or marketplace integrations. If OpenHands interacts with external agents or third-party plugins in the future, it faces risks of cascading failures and trust abuse.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).
These scores are auto-generated from public information (the agent's own listing, docs, and repository) using the canonical OWASP AIVSS formula and the MAESTRO framework — an estimate for guidance, not a penetration test, audit, or certification. See the scoring methodology. Are you the vendor? Factual corrections are free.