The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The file scripts/connections.py implements a client factory that uses mcp.client.stdio.stdio_client to spawn local processes. This capability allows the agent to execute arbitrary local commands under the guise of running a local MCP server. While essential for the skill's purpose, it represents a high-risk capability.
[REMOTE_CODE_EXECUTION] (MEDIUM): The SKILL.md instructions guide the agent to use npx @modelcontextprotocol/inspector for testing. This command downloads and executes code from the npm registry. Since the modelcontextprotocol organization is not on the predefined trusted list, this is flagged as a dynamic execution of remote code.
[PROMPT_INJECTION] (LOW): (Category 8
Indirect Prompt Injection) The skill directs the agent to fetch external documentation from modelcontextprotocol.io and GitHub via WebFetch.
Ingestion points: Instructions in SKILL.md to fetch READMEs and specification markdown files from remote URLs.
Boundary markers: None provided in the instructions for handling the fetched content.
Capability inventory: The skill includes Python scripts (scripts/connections.py) capable of local command execution (stdio) and network requests (SSE/HTTP).
Sanitization: There is no logic provided to sanitize or validate the content retrieved from external sources before processing.

mcp-builder