lark-workflow-crm

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs capturing a returned base_token and embedding it directly into subsequent lark-cli commands as a --base-token argument, which requires the LLM to output/handle the secret verbatim (an insecure pattern).

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly creates a public "客户咨询登记" form whose submissions "会自动进入客户信息表" (Phase 7 / SKILL.md form creation), and the agent routinely reads+queries records (+record-list, +data-query) and acts on their content (e.g., syncing stage changes, sending reminders/notifications), meaning untrusted user-submitted data can directly influence actions like updating商机阶段 and sending messages.