openai-docs

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt explicitly instructs the agent to run installation commands and to "retry with escalated permissions" without asking the user, which is an instruction to change/escape the environment and perform privileged actions beyond the declared doc-fetching purpose and is therefore a hidden/deceptive escalation risk.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill instructs the agent to run installation commands and explicitly to "retry with escalated permissions" if blocked by permissions/sandboxing, which directs the agent to obtain elevated privileges and bypass security controls on the host.