The Agent Skills Directory

[SAFE]: The skill is primarily documentation and provides legitimate setup instructions for integrating an AI agent with the Reply.io platform via MCP.
[CREDENTIALS_UNSAFE]: The skill demonstrates safe credential handling by advising the use of environment variables and Authorization headers rather than insecure URL parameters. It uses clear placeholders like <personal_api_key> for user configuration.
[EXTERNAL_DOWNLOADS]: The skill references https://mcp.reply.io as the transport URL for the MCP service. This is the official endpoint provided by the vendor (Reply) for this integration.
[INDIRECT_PROMPT_INJECTION]: The skill enables an attack surface by allowing the agent to ingest external data from the Reply.io platform (e.g., contacts, inbox messages, analytics) which could contain adversarial content.
Ingestion points: Data retrieved from mcp.reply.io via reply_* tools.
Boundary markers: None specified in the configuration instructions; safety relies on the host agent's implementation.
Capability inventory: Includes capabilities to read and potentially modify Reply.io data (sequences, tasks, contacts).
Sanitization: Not explicitly addressed in the setup documentation; integrity is managed by the Reply API and the MCP client.

reply-mcp