skills/ascend-ai-coding/awesome-ascend-skills/external-gitcode-ascend-vllm-tests-failure-analysis/Gen Agent Trust Hub
external-gitcode-ascend-vllm-tests-failure-analysis
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to perform necessary shell operations for developer tasks, including environment setup via sourcing scripts (
set_env.sh), checking hardware status withnpu-smi info, and executing regression tests usingpytest.- [EXTERNAL_DOWNLOADS]: The skill facilitates the download of model weights and testing assets from established services like Hugging Face (includinghf-mirror.com) and ModelScope. It also recommends installing specific Python packages such asrunai-model-streamerto resolve test dependencies.- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface (Category 8) because its primary workflow involves ingesting and analyzing external test files and execution logs. - Ingestion points: The agent is directed to read test files from the
vllm/tests/directory and log output to/tmp/test_output.log. - Boundary markers: No specific delimiters are mandated to isolate untrusted content from the system instructions.
- Capability inventory: The skill enables file system modifications (copying and patching tests) and the execution of the analyzed code via
pytest. - Sanitization: There are no explicit instructions to sanitize or validate the content of the external test scripts before analysis or execution.
Audit Metadata