huawei-cloud-kafka-replica-reassignment
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided Excel files containing topic names. To mitigate potential injection or path traversal attacks via these names,
scripts/stage3_json.pyimplements a strict whitelist ([A-Za-z0-9._-]) and arealpathcontainment check (_path_inside) before using topic names as filenames for the generated JSON artifacts. - [COMMAND_EXECUTION]: The regression suite
evals/run_evals.pyutilizessubprocess.runto execute pipeline scripts for testing purposes. These calls are confined to the developer test suite, using static paths and internal parameters, and are not reachable by external user-uploaded data processed through the Web UI. - [SAFE]: The Flask application in
scripts/app.pymanages user uploads and artifact downloads safely. It uses a dedicated work directory in the system temp folder (overridable via environment variables) and includes defensive measures such as verifying path prefixes before directory deletion inrun_pipeline_thread.
Audit Metadata