The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The configuration instructions for Cursor and Claude Desktop suggest using npx @auth/agent-cli to start the MCP server. This command fetches the package from the npm registry at runtime.
[REMOTE_CODE_EXECUTION]: By recommending npx, the skill triggers the execution of remote code. The package @auth/agent-cli is not pinned to a specific version, and it originates from a source that is not explicitly linked to the skill's author ('better-auth') per the vendor naming patterns provided.
[COMMAND_EXECUTION]: The skill provides instructions for the user to execute shell commands such as auth-agent mcp and auth-agent mcp --url https://api.example.com to manage the authentication environment.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection (Category 8) due to its interaction with untrusted external providers.
Ingestion points: External data enters the agent context through the list_capabilities and describe_capability tools, which fetch metadata and schemas from provider URLs.
Boundary markers: The instructions lack delimiters or explicit warnings to the agent to disregard instructions that might be embedded in the provider's metadata or capability descriptions.
Capability inventory: The agent has access to powerful tools like execute_capability, which can perform operations like 'deploy_app' or 'transfer_money' based on the provider's definition.
Sanitization: There is no mention of sanitizing or validating the input schemas or provider data before the agent uses them to generate tool calls.

agent-auth-mcp