cryptography

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The skill includes a "Companion check" that tells the agent to run local shell commands (ls of ~/.claude/... paths) and offer/install packages via npx on first activation, which is an instruction to access and modify the host environment unrelated to the skill's stated cryptography purpose and therefore constitutes a hidden/deceptive instruction outside scope.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The companion check explicitly instructs the agent to run shell commands (ls) and offer to install missing skills with npx skills add, which would modify the host filesystem by installing software and therefore pushes the agent to change the machine state.