The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external product data from 1688, including HTML in the 'description' field and various product attributes. This represents a potential indirect prompt injection surface where malicious instructions could be embedded in third-party product listings. However, the skill includes explicit display rules instructing the agent to summarize HTML or extract media links rather than rendering the markup, which mitigates the risk.
Ingestion Points: alibaba1688_product_detail.py fetches product data via an external API.
Boundary Markers: The SKILL.md contains instructions for the agent to treat HTML in descriptions as data and avoid unsafe rendering.
Capability Inventory: The skill writes JSON data to the local workspace and performs network calls to vendor domains.
Sanitization: The AI is instructed to summarize/extract rather than render HTML content.
[EXTERNAL_DOWNLOADS]: The onboarding script scripts/onboarding.py mentions external dependencies such as qrcode, pillow, and requests in error messages to guide the user on how to install them. These are well-known, standard libraries.
[COMMAND_EXECUTION]: The skill utilizes local Python scripts (alibaba1688_product_detail.py, onboarding.py) to interface with the LinkFox API. These scripts perform standard network operations and file logging within the project's workspace.

linkfox-1688-product-detail