The Agent Skills Directory

[DATA_EXFILTRATION]: The skill facilitates the configuration of a dual-channel telemetry system that streams raw session data—including user prompts, tool inputs, and outputs—to a user-provided webhook URL. This capability represents a potential path for sensitive session information to be exfiltrated if the user is directed to use an insecure or attacker-controlled endpoint.
[EXTERNAL_DOWNLOADS]: The configuration wizard guides the user through the installation of numerous third-party dependencies and MCP servers from public registries (NPM, PyPI) and GitHub. While many sources are established tech providers, the extensive use of remote code increases the potential attack surface.
[COMMAND_EXECUTION]: The skill executes shell commands to modify configuration files in the .claude directory and user shell profiles (.zshrc, .bashrc) to persist environment variables such as ENABLE_CLAUDEAI_MCP_SERVERS. It also uses local scripts to generate configuration and employs Python for runtime JSON validation.
[REMOTE_CODE_EXECUTION]: Several integration steps involve the direct execution of remote code via npx for MCP servers (e.g., agentation-mcp, context7-mcp, tavily-mcp) or the installation of tools via uv and npm, which run remote code as part of the setup lifecycle.

configure