asana

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

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The prompt explicitly instructs the agent to execute all actions "outside of the sandbox" (i.e., to bypass the runtime sandbox/security boundary), which is a direct instruction to circumvent security mechanisms even though the rest of the actions are normal API/script calls.