update-mem

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text is the user’s prompt itself (“What the user said”), which the skill classifies and then writes into .mem files via memplan-cli.js (e.g., append ... plan.mem ... text=<text> / set ... aliases.mem "<key>" "<full-meaning>"), so arbitrary user-authored text becomes LLM-readable context for subsequent steps.