Software supply chain risk is the chance that a third-party component, update, plugin, or integration has been altered, backdoored, or otherwise compromised before it reaches your system. The danger is not only in code you write yourself; it also comes from the libraries, packages, APIs, and services your software depends on.
In cyber security, this matters because attackers often target the trust chain instead of the final application. A poisoned dependency, malicious update, or compromised MCP server can give an attacker access to data, commands, or credentials without needing to break the main product directly. Defenses focus on inventorying dependencies, verifying signatures, restricting permissions, reviewing updates, and treating external inputs as untrusted until validated. In AI agent environments, the risk grows because plugins and skills can shape both what the model knows and what actions it can take, making supply-chain control a core part of securing the agent.