The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The instructions in SKILL.md suggest the installation of an external skill from a third-party repository (benjitaylor/skills) for specific UI components, which is not managed by the author or a trusted vendor.
[COMMAND_EXECUTION]: The utility script scripts/check.mjs uses spawnSync to execute other scripts (killer-scan.mjs and corta-url.mjs) by navigating to relative sibling directories. While the script names are hardcoded, this represents a pattern of local command execution to perform audits.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to audit user-provided source code and live application URLs, which creates a surface for indirect prompt injection if the processed content contains malicious instructions.
Ingestion points: Local UI source code paths and external application URLs processed by scripts/check.mjs.
Boundary markers: There are no explicit instructions or code logic to use delimiters or ignore embedded instructions within the processed UI content or screenshots.
Capability inventory: The skill has the ability to execute subprocesses through scripts/check.mjs and read file system/network data via audit tools.
Sanitization: The skill lacks clear sanitization or validation logic for the content it ingests from external URLs or local source files before analysis.

visant-design