The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes system binaries including sips, cwebp, and ImageMagick's convert tool using child_process.spawn. It passes parameters as an argument array, which is a secure practice to prevent shell injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied file paths and directory names, creating a potential surface for indirect injection if metadata is handled insecurely.
Ingestion points: CLI input arguments in scripts/main.ts and settings parsed from EXTEND.md files.
Boundary markers: No specific delimiters or instructions are provided to the agent to distinguish between data and instructions in these inputs.
Capability inventory: The skill can execute subprocesses via spawn and perform file system operations such as renameSync and unlinkSync across the workspace.
Sanitization: The script utilizes path.resolve and path.join for path normalization but lacks comprehensive sanitization for file names or external configuration values.
[DYNAMIC_EXECUTION]: The script uses dynamic import() to load the sharp library at runtime. While the import targets a static package name, dynamic loading is a vector for potential environment-based execution issues.

tuzi-compress-image