Agent data injection
Named by 2 kept items in this workspace.
Search every workspace for Agent data injection. It matches the name as written, so it can find other things that share it.
The items that name it 2
Hides malicious instructions in data disguised as trusted metadata or context, slipping past indirect prompt injection defenses; used to spoof pull request authorship and trigger command execution
Indirect prompt injection class hiding malicious instructions inside data disguised as trusted metadata or context; demonstrated against Claude Code, Codex, Gemini CLI, Claude in Chrome (paper submitted July 6, 2026)
All techniques in this workspace.