triton-ascend-migration
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill instructions explicitly direct the agent to generate Python verification scripts and execute them (e.g.,
python verify_kernel.py) to validate migration results. This involves the generation and execution of code at runtime based on patterns defined in the skill. - [COMMAND_EXECUTION]: The skill provides templates for shell commands, including environment variable exports (
export ASCEND_RT_VISIBLE_DEVICES) and container interaction (docker exec), which the agent is expected to use to perform and verify migrations. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code snippets and file content provided by the user. Since this data is used to generate scripts that the agent then executes, it presents a surface for indirect prompt injection.
- Ingestion points: User-pasted code snippets and file paths specified in queries.
- Boundary markers: The instructions lack explicit delimitation or safety warnings when interpolating user code into analysis and script generation tasks.
- Capability inventory: The agent has shell execution (
python,bash,docker) and file-writing capabilities across the migration workflow. - Sanitization: There is no explicit requirement to sanitize or validate user-provided code before it is used to construct executable verification scripts.
Audit Metadata