The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions as a code-generation and documentation assistant that ingests external content from the local repository (e.g., tests/docs/python/test_base.py and tests/docs/c/test_base.cpp).
Ingestion points: The agent is instructed to resolve inputs from existing repository files and specific referenced snippets.
Capability inventory: The skill utilizes the Read and Grep tools to access and interpret the codebase.
Boundary markers: There are no explicit delimiters defined to separate ingested file content from the agent's internal reasoning or system prompt.
Sanitization: No specific validation or filtering of the ingested content is documented.
Note: This is an inherent surface for Retrieval-Augmented Generation (RAG) skills; since the references are restricted to the vendor's own repository structure, this surface is considered within the expected scope of the skill's primary purpose.
[SAFE]: The skill follows secure metadata practices. The author and license information align with the provided NVIDIA context. No obfuscation, unauthorized network operations, or persistence mechanisms were detected. The technical advice provided (such as stepping the renderer for 40 frames) is legitimate domain-specific knowledge for rendering software.

warmup