ziniao-device

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes CLI examples that pass a plaintext password flag (e.g., --proxy-pass secret), which requires the agent to include user-provided secret values verbatim in commands or code, creating an exfiltration risk.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill explicitly requires at runtime that the agent "MUST 先用 Read 工具读取 ../ziniao-shared/SKILL.md", which is a required external file read that can alter agent instructions/prompts (flagged: ../ziniao-shared/SKILL.md).

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). Yes. The skill exposes explicit purchase/renewal operations that perform real charges: commands "device purchase", "device renew", and "device auto-renew" are marked as involving fees/high-risk-write. Underlying API endpoints include /superbrowser/rest/v1/erp/purchase, /superbrowser/rest/v1/erp/renew, /superbrowser/rest/v1/erp/package/new, and /superbrowser/rest/v1/erp/ip/renewal. These are specific write APIs to initiate billing/purchase actions (i.e., move money or trigger charges), so the skill grants Direct Financial Execution authority.