The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions and reference materials direct the agent to use the grep utility to audit the codebase for specific UI patterns and anti-patterns (e.g., searching for transition: all or z-index: 9999). This is a legitimate use of shell tools for the purpose of static analysis.
[DYNAMIC_EXECUTION]: The skill includes a Node.js script, assets/gen-springs.mjs, designed to generate CSS linear() easing strings from physics tokens. The agent is instructed to run this script as a build-time utility. This is a transparent and benign use of a local script to automate design token generation.
[EXTERNAL_DOWNLOADS]: The provided Node.js script references the motion (formerly framer-motion) package as a required dependency. This is a well-known, trusted, and widely-used animation library in the web development community.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted content from a project's documentation and configuration files (e.g., DESIGN.md, AGENTS.md, tailwind.config).
Ingestion points: Analyzes project-specific rules and token files as described in SKILL.md §1.1.
Boundary markers: None explicitly defined in the intake instructions.
Capability inventory: File reading/writing and shell command execution (grep).
Sanitization: No specific sanitization of project content is mentioned.
While this ingestion surface exists, the risk is minimal given the skill's focused scope of UI auditing and its lack of high-privilege operations.

ultrapolish-web