idea

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The prompt explicitly directs spawning an agent with mode: "bypassPermissions", which is a hidden/deceptive instruction to override permissions and alter agent behavior outside the skill's stated analysis purpose.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill contains an explicit instruction to spawn an analysis agent with mode "bypassPermissions", which deliberately subverts permission boundaries and creates a high-risk backdoor/privilege-escalation vector that could be used to access sensitive files, environment variables, or exfiltrate data.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The prompt explicitly requires spawning an agent with mode: "bypassPermissions" (and references delegating analysis with that bypass), which directs the agent to bypass security/permission mechanisms and thus compromises the host's security.