The Agent Skills Directory

[SAFE]: The skill is purely instructional and provides guidance on Linux command-line operations without executing them directly on the user's system.
[SAFE]: Includes a dedicated 'Security Warning List' that identifies high-risk commands (e.g., rm -rf /, chmod 777, fork bombs, and wget | bash) and provides safe alternatives or warnings against their use. The presence of these patterns is for educational and defensive purposes.
[SAFE]: The 'Security First' core principle instructs the agent to warn users about dangerous operations, prioritize minimal privilege, and explain parameters to ensure the user understands what they are running.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a command and script generator based on untrusted user input (Step 1: Understand demand). While this creates an attack surface for indirect injection, the risk is mitigated by the skill's instructions to provide explanations, safety warnings, and the requirement for manual user execution of the generated code. Evidence: Ingestion point in 'Step 1: 理解需求'; Boundary markers include markdown code blocks in '输出格式'; Capability inventory includes shell script and command generation; Sanitization consists of parameter explanations and security warnings. Severity: SAFE.

linux-helper