The Agent Skills Directory

[SAFE]: The skill provides clear instructions for interacting with the text_to_avatar tool within the @three-ws/mcp-server environment. It outlines expected inputs, outputs, and the payment flow using the x402 protocol.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text prompts and image URLs to generate 3D models. While this represents a data ingestion surface, the risk is inherent to the primary function of the skill (text-to-3D) and does not grant the agent dangerous capabilities.
Ingestion points: prompt and images parameters in the text_to_avatar tool call (SKILL.md).
Boundary markers: None explicitly defined in the instructions.
Capability inventory: Limited to calling the specified MCP tool and returning a URL. No file writing or shell execution capabilities are utilized.
Sanitization: Not explicitly mentioned; expected to be handled by the backend service (Replicate/Hunyuan-3D).
[COMMAND_EXECUTION]: No shell commands, system calls, or script execution patterns were detected in the skill instructions.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or sensitive environment variable paths are present. The payment address is handled by the platform transport, not hardcoded in the skill.

text-to-avatar