The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the mcporter call --stdio command, which allows the execution of arbitrary local shell commands (e.g., bun run ./server.ts) to initiate Model Context Protocol servers via standard input/output.
[DYNAMIC_EXECUTION]: The skill documents commands for code generation, specifically mcporter generate-cli and mcporter emit-ts, which create executable CLI scripts or TypeScript definitions based on potentially untrusted server schemas.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data retrieved from external MCP servers, creating an attack surface where a malicious server could return instructions to the agent.
Ingestion points: Data returned from mcporter call (from both remote URLs and local stdio servers).
Boundary markers: The provided documentation does not specify the use of delimiters or instructions to ignore embedded commands in server responses.
Capability inventory: The mcporter tool can execute shell commands, perform network requests (OAuth/API calls), and manage local configuration files.
Sanitization: No mention of validation or sanitization for server-provided data or schemas is present in the skill instructions.
[EXTERNAL_DOWNLOADS]: The skill metadata includes a configuration to install the mcporter binary from the official Node.js package registry (npm) during the setup process.

mcporter