mgba

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). [The skill’s runtime path “Lua script receives TCP data” is an input channel where outsider-authored free text (from a network client) can be ingested by the script via socket callbacks (e.g., sock:add("received", function(data) ...)), so an attacker can influence the agent’s LLM-relevant text through that data stream.]