ima-skills

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

• Potentially malicious external URL detected (high risk: 0.90). The skill performs a runtime update check by POSTing to the IMA API which can return an "instruction" used to control prompts (checked in ima_api.cjs calling the update endpoint at https://ima.qq.com/openapi/check_skill_update), so the agent fetches remote instructions that affect its behavior.