The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted product descriptions, names, and attribute data retrieved from external Shopee listings, which could be used to influence the agent's behavior.
Ingestion points: Product details (description, name, attributes) are ingested by scripts/shopee_product_detail.py from the Shopee marketplace via the NexScope proxy.
Boundary markers: Absent. The skill instructions in SKILL.md provide formatting guidelines but do not include explicit instructions for the agent to ignore or delimit potentially malicious text within the product description.
Capability inventory: The script scripts/shopee_product_detail.py has the ability to perform network requests, write to the local filesystem (caching and output logging), and access environment variables (NEXSCOPE_API_KEY).
Sanitization: The validate_params function in scripts/shopee_product_detail.py strictly validates the productUrl against supported Shopee domains and a specific URL pattern before execution.
[COMMAND_EXECUTION]: The skill instructions and documentation encourage the execution of a local Python script (scripts/shopee_product_detail.py) to interface with the NexScope API. This is the primary intended mechanism of the skill.
[DATA_EXPOSURE]: The script scripts/shopee_product_detail.py automatically writes the full, raw API response to the local filesystem (under nexscope/ in the current working directory or home folder). While SKILL.md contains a privacy notice to redact credentials from user-facing output, the script itself persists the raw data, which could contain shop-specific or session-specific metadata.

ecommerce-shopee-product-detail