Missing User Warnings
Medium
- Confidence
- 89% confidence
- Finding
- The script appends untrusted input directly into persistent memory or quarantine files with no interactive confirmation, provenance hardening, or content-safe encoding beyond truncation/whitespace cleanup. In this skill's context, that is meaningful because the tool is specifically intended to protect memory from prompt-injection and data-exfiltration content, yet it still persists attacker-controlled text, including full original text in quarantine, which can poison later workflows or preserve sensitive content for future retrieval.
