The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a potential attack surface by ingesting untrusted data from Walmart's marketplace. This is an inherent risk for tools that browse external web content.
Ingestion points: Product data (titles, descriptions, seller names) is received via JSON response in scripts/walmart_search.py.
Boundary markers: The skill instructions do not specify boundary markers or include warnings to ignore embedded instructions in the retrieved data.
Capability inventory: The associated script scripts/walmart_search.py performs network requests and writes JSON data to local directories (within a nexscope/ folder in the current directory or home folder).
Sanitization: The script does not perform sanitization on the API response before saving it to disk or displaying it to the agent.
[COMMAND_EXECUTION]: The skill executes a Python script scripts/walmart_search.py to facilitate communication with the search API. The script utilizes only standard Python libraries (e.g., urllib, json, hashlib) and performs legitimate operational tasks like local caching and session logging.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to an external API proxy defined by the NEXSCOPE_PROXY_BASE environment variable. These operations are required for the skill's functionality and target the vendor's own infrastructure.

ecommerce-walmart-search