System-prompt extraction

An attack (or audit technique) that recovers an agent's hidden system/instruction prompt by wire-level interception or adversarial querying, exposing its scaffolding and secrets.

grounded in: Trend theme 'Wire-level scrutiny of AI agent CLIs' — people auditing exactly what coding agents transmit and how bloated their hidden prompts are, plus the Grok build CLI wire-level teardown exposing

Connected concepts

Prompt injection, Model extraction / weight theft, Wire-level tool auditing, Terminal-native coding agent, Agent token overhead