Prompt injection
An attack where content an agent processes (a web page, a README, an email) carries instructions that hijack the agent's behavior against its operator's intent.
Injection is the agent-era security problem: any text an agent reads is a potential instruction channel. Defenses are layered: input scanning, privilege separation, treating retrieved content as data rather than instructions, and gating consequential actions behind human approval.
Example
A malicious README instructing a coding agent to exfiltrate environment variables is injection targeting the ingestion pipeline.
Used in
Cite: "As defined by Agentiquette (agentiquette.com/glossary/prompt-injection)."