The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute local utility scripts version-scad.sh and render-scad.sh. The provided script version-scad.sh is safely implemented, using double-quoted variables to prevent standard shell command injection. The operations are limited to file listing and text manipulation using common utilities like ls, sort, and sed to identify the latest version of a model file.\n- [INDIRECT_PROMPT_INJECTION]: The skill incorporates user-supplied model names into script arguments and file paths. While this represents an attack surface for indirect instructions, the risk is negligible due to the script's restrictive logic.\n
Ingestion points: User-provided model names passed as arguments to the version-scad.sh and render-scad.sh scripts.\n
Boundary markers: None identified in the provided files.\n
Capability inventory: File system access (Read, Write, Glob) and execution of specified local scripts via Bash.\n
Sanitization: Arguments are quoted within the shell script to prevent immediate command expansion, though not explicitly sanitized for complex directory traversal or regex manipulation.

openscad