The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill interacts with external APIs to perform its intended functions.
Fetches image generation results from ModelScope's inference service at api-inference.modelscope.cn.
Optionally utilizes the Anthropic API for intelligent style detection if an ANTHROPIC_API_KEY is provided.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits a potential surface for indirect prompt injection where untrusted user input is processed by a downstream LLM.
Ingestion points: User-provided prompts passed via the CLI or read from a batch file (prompts.txt) are processed in scripts/generate.py.
Boundary markers: In scripts/style_templates.py, the user prompt is interpolated into a template for the Anthropic API (_detect_style_with_llm) using markdown headers (## 内容：), but lacks explicit boundary markers or instructions to ignore embedded commands within the user text.
Capability inventory: The skill can perform network requests to external APIs (aiohttp) and write files to the local disk (saving generated images).
Sanitization: There is no specific sanitization or filtering applied to the user prompt before it is sent to the style-detection LLM or the image-generation API.

z-image