The Agent Skills Directory

[COMMAND_EXECUTION]: The font optimization script provided in references/font-management.md utilizes child_process.exec to run external binaries. The input variable, containing file paths derived from fs.readdir, is interpolated directly into the shell command strings (woff2_compress ${input} and sfnt2woff ${input}) without any escaping or sanitization. This pattern allows for command injection if a file in the input directory has a name containing shell metacharacters.
[INDIRECT_PROMPT_INJECTION]: The skill implements an automated asset processing pipeline that ingests data from the local file system, creating an attack surface for indirect injection via malicious file naming.
Ingestion points: references/font-management.md, references/image-optimization.md, and references/organization.md all use fs.readdir to read files from the project directory.
Boundary markers: None are present to delimit or validate the filenames before processing.
Capability inventory: The skill uses child_process.exec for font compression, fs.rename for file organization, and fs.writeFile for generating optimized versions and manifest files.
Sanitization: Filenames are used directly in shell commands and file system operations without sanitization or validation against allowed patterns.
[DYNAMIC_EXECUTION]: The skill provides logic to dynamically invoke command-line tools (woff2_compress, sfnt2woff) at runtime based on the contents of the asset directories, which facilitates the execution of code outside the primary agent environment.

asset-manager