add-jit-kernel
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides a legitimate technical tutorial for adding JIT CUDA kernels to the sglang project. All code snippets follow standard development practices for high-performance computing libraries.\n- [DYNAMIC_EXECUTION]: The skill demonstrates the use of a Just-In-Time (JIT) compilation system (load_jit) to build and execute CUDA kernels. This is a primary feature of the framework and is used appropriately within the project's infrastructure.\n- [INDIRECT_PROMPT_INJECTION]: While the skill defines APIs that ingest external tensor data, it utilizes robust validation through TensorMatcher in the C++ layer to enforce type safety, alignment, and device constraints, mitigating common attack vectors at this level.
Audit Metadata