The Agent Skills Directory

[SAFE]: The skill provides instructions for using standard command-line tools such as find, rg (ripgrep), and cmake to navigate and build source code within a local checkout of the CUTLASS library. These operations are restricted to the local filesystem and serve the skill's primary purpose of GPU kernel development.
[SAFE]: The skill references local automation scripts (update-repos.sh, install.sh, validate_repo.py) for maintaining the repository state. These scripts are part of the skill's operational framework provided by the author.
[INDIRECT_PROMPT_INJECTION]: The skill involves reading and analyzing external source code and examples from the CUTLASS repository, which constitutes an ingestion point for untrusted data.
Ingestion points: The agent reads C++, Python, and configuration files from the $CUTLASS_REPO directory.
Boundary markers: The instructions do not specify explicit delimiters or markers to isolate instructions found within the source code.
Capability inventory: The skill utilizes find, ripgrep, and cmake for search and compilation tasks.
Sanitization: There is no explicit sanitization of the content read from the external repository before processing.

cutlass-skill