skills/ascend-ai-coding/awesome-ascend-skills/external-gitcode-ascend-ascendc-operator-performance-eval/Gen Agent Trust Hub
external-gitcode-ascend-ascendc-operator-performance-eval
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions require the agent to read and process content from external documentation files (e.g., design.md and op-test-cases.md) to generate test cases. While this creates a surface for indirect prompt injection if those files were maliciously crafted, it is a functional requirement for the skill's automation goals. Mandatory Evidence Chain: (1) Ingestion points: csrc/ops//design.md and csrc/ops//test/-test-cases.md. (2) Boundary markers: Absent. (3) Capability inventory: File system read/write, execution of Python benchmarking scripts. (4) Sanitization: Absent.
- [DYNAMIC_EXECUTION]: The provided reference implementation (layer_norm_profiler_common.py) uses torch.ops.load_library to dynamically load compiled operator libraries (.so files). This is standard practice for testing custom kernels in the Ascend hardware ecosystem and occurs within the project's local directory structure.
- [COMMAND_EXECUTION]: The skill guides the user and the agent through executing Python benchmarking scripts that interact with hardware profilers and the filesystem. These actions are within the scope of the skill's primary function as a performance evaluation tool.
Audit Metadata