The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-provided prompts for media generation. While this constitutes an indirect prompt injection surface, the skill includes explicit safety instructions to rephrase prompts upon safety rejection and to refuse requests for identifiable likenesses.
[CREDENTIALS_UNSAFE]: The skill follows security best practices by explicitly instructing the agent not to hardcode API keys in artifacts and instead rely on the local environment configuration of the asset-router MCP.
[COMMAND_EXECUTION]: The skill interacts with external providers through a local MCP server using standard tool calls. It does not perform arbitrary command execution or unsafe subprocess spawning.
[EXTERNAL_DOWNLOADS]: The skill references well-known media generation services (Adobe, ElevenLabs, etc.) for its primary functionality. It does not download or execute unverifiable external code; all external interactions are mediated by the referenced MCP server.

asset-generator