customer-research

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The "Companion check" instructs the agent to run local shell commands (ls ~/.claude/skills/ ...) and offer installing packages via npx, which is an explicit, out-of-scope directive that accesses the host environment and can exfiltrate or modify system state rather than performing customer research.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs the agent at runtime to run "npx skills add AbsolutelySkilled/AbsolutelySkilled --skill " (and the specific install "npx skills add AbsolutelySkilled/AbsolutelySkilled --skill customer-research"), which uses npx to fetch and execute remote package code during runtime.