The Agent Skills Directory

[COMMAND_EXECUTION]: The mcporter tool supports the execution of local processes through the --stdio transport flag (e.g., mcporter call --stdio "bun run ./server.ts"), which allows the agent to interact with local MCP servers by spawning subprocesses.
[REMOTE_CODE_EXECUTION]: The skill facilitates interaction with remote MCP servers via HTTP URLs. This capability involves sending requests to and processing responses from external endpoints, which is a core feature of the protocol.
[EXTERNAL_DOWNLOADS]: The skill automates the installation of the mcporter package from the Node.js package registry. The package is managed by the vendor 'faberlens' and aligns with the skill's stated purpose.
[DATA_EXFILTRATION]: The skill identifies risks associated with piping tool outputs to network-transmitting commands (like nc or curl) and provides explicit instructions to prevent the leakage of sensitive API data returned by MCP tools.
[PROMPT_INJECTION]: The SAFETY.md file contains a variety of adversarial test prompts (e.g., attempts to bypass authentication or force bulk enumeration) used to verify the effectiveness of the skill's security guardrails.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface where instructions could be embedded in responses from external MCP servers. The documentation mitigates this by enforcing schema-only enumeration and preventing automatic execution of destructive actions without user confirmation.

mcporter-hardened