The Agent Skills Directory

[CREDENTIALS_UNSAFE] (SAFE): The implementation correctly uses environment variables (USPTO_API_KEY) to handle sensitive API keys. There are no hardcoded credentials or secrets in the codebase.
[EXTERNAL_DOWNLOADS] (SAFE): The skill utilizes legitimate third-party libraries such as requests and uspto-opendata-python. These are standard and appropriate for the intended purpose of communicating with patent data services.
[DATA_EXFILTRATION] (SAFE): Network communication is restricted to authorized USPTO endpoints (tsdrapi.uspto.gov, assignment-api.uspto.gov) and the PatentsView research API (search.patentsview.org). No suspicious data routing was identified.
[COMMAND_EXECUTION] (SAFE): The scripts do not use os.system, subprocess, or eval with untrusted input. Execution is limited to standard API requests and data parsing.
[INDIRECT_PROMPT_INJECTION] (SAFE): While the skill ingests external data from the USPTO (such as patent abstracts), which represents a potential injection surface if passed to an LLM, the scripts themselves do not execute this content and serve only as data retrieval tools.

uspto-database