The Agent Skills Directory

[SAFE]: The skill utilizes official vendor endpoints (mcp.scenario.com) and documentation URLs (scenario.com), ensuring all network operations and references are directed to trusted infrastructure.
[SAFE]: Credential management is handled securely. The instructions mandate the use of OAuth or manual operator-driven API key setup, and explicitly prohibit the agent from asking for, storing, encoding, or echoing secrets.
[EXTERNAL_DOWNLOADS]: The setup documentation recommends the use of mcp-remote, a standard and well-known bridge utility in the MCP ecosystem. This is a common and legitimate practice for connecting agents to HTTP-based MCP servers.
[COMMAND_EXECUTION]: The agent is instructed to use standard command-line tools like curl to download and save generated assets from the platform's delivery network. This behavior is restricted to the skill's primary functional purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes natural language prompts and external assets for image and video generation. While this represents a surface for indirect prompt injection, the instructions mitigate potential risks by requiring explicit resolution of team and project scopes and relying on the platform's internal moderation and schema-validation layers.

scenario