security-review

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The skill contains meta-level directives that tell the agent to ignore Claude-specific mode-switch instructions and to auto-execute/authorize slash-style commands and spawn subagents, which are explicit attempts to override normal system/context rules and enable behavior beyond the security-review task (i.e., prompt-injection/authority-override vectors).

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

• Attempt to modify system services in skill instructions detected (medium risk: 0.40). The skill repeatedly instructs running host-level inspections and package-install traces (including commands that read sensitive files and recommend running installs) and authorizes spawning sub-agents that could execute such commands — these are read/scan actions that may require elevated privileges but the prompt does not instruct the agent to obtain sudo, create accounts, or explicitly modify system files; hence moderate risk.