skills/ascend-ai-coding/awesome-ascend-skills/external-gitcode-ascend-catlass-operator-performance-optim/Gen Agent Trust Hub
external-gitcode-ascend-catlass-operator-performance-optim
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute compilation and profiling commands using local tools like ascendc-operator-compile-debug and ascendc-operator-performance-eval. These are necessary for the skill's primary function of performance tuning in the Ascend development environment.
- [DATA_EXFILTRATION]: The skill requires reading project files, including operator source code, tiling configurations, and profiler results within the local project root. While this involves accessing sensitive intellectual property, the access is confined to the local environment for comparison and reporting.
- [PROMPT_INJECTION]: The skill ingest data from Markdown optimization guides and local profiler output, creating a potential surface for indirect prompt injection.
- Ingestion points: catlass/docs/1_Practice/10_matmul_optimization.md and _torch_npu_profiler_report.md.
- Boundary markers: The skill does not provide explicit boundary markers or instructions to ignore embedded commands within these files.
- Capability inventory: The agent has capabilities to modify source code, recompile binary files, and execute profiling scripts.
- Sanitization: No explicit sanitization of the content of the Markdown files or profiler reports is mentioned.
Audit Metadata