The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes a utility script scripts/fetch_api_support.py designed to scrape the official cuPyNumeric documentation at nvidia.com to refresh its local API support manifest. This is a legitimate functional requirement to ensure migration assessments are based on the latest vendor data.
[DATA_EXFILTRATION]: No unauthorized data exfiltration patterns were found. The skill instructions explicitly direct the agent to perform a read-only assessment and prohibit printing or transmitting environment variables, secrets, or user data externally.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading untrusted user source code using tools like Read and Grep. This presents an indirect prompt injection surface. The skill mitigates this by instructing the agent to perform a read-only semantic review and explicitly forbidding the execution of user code.
[COMMAND_EXECUTION]: The skill documentation suggests specific CLI commands for the user to run (e.g., legate and cuPyNumeric Doctor) but does not invoke shell execution itself.
[SAFE]: The automated scanner alert for numpy.info is a false positive. The string appears in assets/api-support.md as a standard NumPy API name in a comparison table and does not function as a malicious domain in this context.

cupynumeric-migration-readiness