The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The troubleshooting section suggests applying chmod 777 to the log directory (/var/log/mcp-gateway). This practice is a security risk as it grants read, write, and execute permissions to all users on the system, potentially exposing sensitive data or allowing malicious modification of log files.
[PRIVILEGE_ESCALATION]: The instructions require mounting the host Docker socket (/var/run/docker.sock) into the gateway container. This provides the container with full administrative access to the host's Docker daemon, which can be used to escape the container and compromise the host machine.
[COMMAND_EXECUTION]: The core functionality of the gateway involves executing shell commands and launching Docker containers based on user-defined configurations (TOML/JSON). This allows for dynamic execution of potentially untrusted commands if the configuration files are sourced from unauthorized parties.
[INDIRECT_PROMPT_INJECTION]: The skill's role as an MCP gateway creates a vulnerability surface for indirect prompt injection through untrusted data ingestion.
Ingestion points: Backend MCP server responses and client JSON-RPC requests entering via HTTP endpoints.
Boundary markers: None identified for distinguishing external data from agent instructions.
Capability inventory: Significant capabilities including host file system access (via mounts) and arbitrary command execution (via Docker).
Sanitization: Employs schema normalization and configuration validation rules to mitigate structural risks.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download and execution of containerized tools from the GitHub Container Registry (ghcr.io/github/*). These resources originate from a well-known and trusted organization.

gh-aw-mcp-gateway