tilelang-ascend-kernel
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates environment setup and testing through local shell commands. It instructs the agent to set environment variables (
export TL_ROOT), source local initialization scripts (source set_env.sh), and execute Python modules for compilation and testing (python -m compiler...,python setup.py test). These operations are standard for development workflows and target files within the local project repository. - [PROMPT_INJECTION]: The skill involves processing and reviewing source code files, which presents a surface for indirect prompt injection via malicious code comments.
- Ingestion points: Kernel source files and build artifacts (e.g.,
xllm/compiler/tilelang/targets/ascend/kernels/*.py,manifest.json) mentioned in the skill files. - Boundary markers: None present; the instructions do not specify delimiters to separate untrusted source content from instructions.
- Capability inventory: Ability to execute shell commands, Python scripts, and build tools within the terminal.
- Sanitization: No content validation or sanitization is performed on the files before they are processed by the agent.
Audit Metadata