The Agent Skills Directory

COMMAND_EXECUTION (LOW): The core functionality of the skill (found in scripts/mcp-client.ts) involves spawning subprocesses via StdioClientTransport. This is the intended primary purpose of the skill to run MCP servers. The commands and arguments are sourced from a local configuration file (.claude/.mcp.json).
EXTERNAL_DOWNLOADS (LOW): The skill documentation (README.md, references/gemini-cli-integration.md) recommends installing the gemini-cli (a tool from the trusted google-gemini organization) and various MCP servers using npm and npx. While these involve external downloads, they are standard components of the MCP ecosystem.
PROMPT_INJECTION (LOW): The skill is susceptible to indirect prompt injection because it fetches and processes data from external MCP servers without applying strict boundary markers or sanitization.
Ingestion points: Data is ingested through getAllTools, getAllPrompts, and getAllResources methods in scripts/mcp-client.ts which communicate with external servers.
Boundary markers: No specific delimiters or security instructions are used when presenting the output of MCP servers to the agent.
Capability inventory: The skill can execute arbitrary commands (configured as MCP servers) and perform file-write operations (assets/tools.json).
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from external MCP servers.

mcp-management