The Agent Skills Directory

[Command Execution] (LOW): The skill instructs the agent to execute a Python script using user input as a command-line argument (python scripts/detect_mcp.py "用户查询内容"). This creates a potential risk of command injection if the agent's execution environment fails to properly escape shell metacharacters (such as ;, &, or |) provided within the user query.- [Indirect Prompt Injection] (LOW): The skill creates a surface for indirect prompt injection by processing untrusted data to influence tool selection.
Ingestion points: User queries are ingested via sys.stdin or sys.argv in scripts/detect_mcp.py.
Boundary markers: None. The script processes raw strings without delimiters or instructions to ignore embedded commands.
Capability inventory: The script produces JSON recommendations that guide the agent's decision on which MCP tools (like web_search or grep_app) to invoke next.
Sanitization: The script performs minimal cleaning by removing markdown code blocks, but does not validate or sanitize input to prevent users from intentionally triggering specific tool recommendations via keyword stuffing.

mcp-detector