The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the mcpjam CLI tool to perform diagnostics, including server probing (server probe), tool execution (tools call), and OAuth conformance testing (oauth conformance).
[EXTERNAL_DOWNLOADS]: The skill interacts with external MCP servers and OAuth authorization endpoints at URLs provided by the user to perform security reviews and capability discovery.
[PROMPT_INJECTION]: The skill processes and interprets data from external, untrusted MCP servers. It incorporates a "Default stance" and "Hard rules" requiring the agent to prioritize raw request/response evidence and interpret findings conservatively to mitigate the risk of indirect prompt injection from malicious server outputs.
[SAFE]: The skill manages sensitive OAuth tokens by saving them to local files (--credentials-out). It explicitly instructs the user/agent to use secure file permissions (mode 0600) and notes that debug outputs (--debug-out) automatically redact secrets, following security best practices for local credential management.

mcp-inspector