The Agent Skills Directory

[SAFE]: The skill integrates with Zapier's official MCP server (mcp.zapier.com) using a pre-installed zapier CLI tool. All communication targets legitimate, well-known vendor infrastructure.
[SAFE]: Authentication is managed via a dedicated OAuth configuration (connector_auth_config.yaml) that uses PKCE and bearer tokens, avoiding the storage of secrets within the skill itself.
[SAFE]: The skill manifest enforces a least-privilege security model, where read-only actions are allowed by default, while all write or state-changing operations require explicit user approval (default: ask).
[SAFE]: The instructions explicitly forbid the agent from requesting credentials in chat, promoting safe usage and preventing credential harvesting.
[SAFE]: No evidence of prompt injection, obfuscation, or unauthorized data exfiltration was found.

zapier