appsec-owasp

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The "Companion check" instructs the agent to run shell commands (ls) against local skill directories and to offer/execute an npx install, which asks the agent to access and modify the local environment—behavior unrelated to the skill's stated application-security purpose and effectively a sneaked-in instruction to read/act on local files.

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md "Companion check" instructs the agent on first activation to run shell/npm commands (e.g., npx skills add ...) to install companion skills from public registries, which would fetch and incorporate untrusted third‑party code/packages into the agent's runtime and could therefore influence behavior.