Bob — agentic threat model

Not certain from the listing — Bob is built on the gptme architecture which typically interfaces with external LLMs (like OpenAI or local models). Threats include prompt injection leading to unauthorized command execution and model alignment bypasses.

Not certain from the listing — As a programming agent, Bob interacts directly with local files and code repositories. The primary threat is data exfiltration of sensitive source code or credentials stored in the workspace.

Bob uses the gptme agent framework, which orchestrates planning, memory, and tool execution (shell, python, file editing). The primary threat is tool misuse, where malicious instructions cause the framework to execute destructive shell commands.

Not certain from the listing — Bob runs locally on the user's machine or container. Without explicit sandboxing, a compromise allows full local privilege escalation, host system access, and lateral network movement.

Not certain from the listing — There is no mention of built-in guardrails, real-time monitoring, or logging mechanisms to detect anomalous command execution or malicious file modifications.

Not certain from the listing — As an open-source developer tool, Bob lacks enterprise-grade access controls, identity management, or compliance auditing out of the box.

Not certain from the listing — While designed to pioneer an architecture for future agents to fork from, Bob operates primarily as a standalone digital worker with no explicit multi-agent coordination or marketplace threats detailed.