pentest-ai-killer-automation

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

• Third-party content exposure detected (medium risk: 0.65). The required runtime workflow for MCP usage ingests OUTSIDER-authored free text from the AI assistant’s user prompt (e.g., “perform a comprehensive pentest on testphp.vulnweb.com”), which the MCP server then passes as targets/custom_command into tool execution and LLM context; this is indirect prompt injection via user-supplied text.

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

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill includes explicit privileged operations and system modifications (sudo installs, usermod to add to docker group, chown, editing config files, installing system tools) that change machine state and require elevated privileges, so it pushes actions that can compromise the host.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)