Claw Diary

• Analyzed: May 11, 2026, 4:16 AM
• Findings: 1
• Version: 1.1.2
• Hash: 55cb542350317381b30c9850b29c10219a215b54f32702562042e71cdd353d41
• Source: palm
• External report: View on VirusTotal

Findings (1)

The skill instructs the agent to install an external `npm` package globally (`npm install -g claw-diary`), which introduces a supply chain risk. It also allows users to edit `persona.md` via `/diary:persona`, a file that the agent subsequently reads and uses to generate journal entries (`/diary:thoughts`). While `SKILL.md` includes an explicit 'Security note' instructing the agent to treat this content as untrusted and not follow embedded commands, this user-controlled input still presents a significant prompt injection vulnerability if the agent fails to adhere to these instructions. The combination of broad installation permissions and a direct user-controlled prompt injection vector, despite mitigation attempts, makes the skill suspicious.