AI coding agents like Claude Code, Gemini CLI, and OpenCode can write exploits and reason about targets, but they have no clean way to interact with a reverse shell. agent2shell bridges that gap.It catches reverse shell connections over TCP and exposes them as structured JSON APIs via Unix domain sockets. Any process on the operator machine (your AI agent) can execute commands, transfer files, query session metadata, and broadcast across multiple sessions. You get clean output, exit codes, and timeouts.It ships with skill files that teach AI agents how to use it. Drop the skill into the agent's config directory and it can autonomously run commands on compromised hosts, push/pull files, pipe scripts without touching disk (--stdin), and manage multi-session engagements.Single static Go binary with zero runtime dependencies. Not a C2 framework, not an AI pentester. Just the infrastructure layer between them. Comments URL: https://news.ycombinator.com/item?id=48009900 Points: 1 # Comments:...