The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates downloading code and design system configuration from an external, non-trusted source.
Evidence: Points to https://github.com/ryanthedev/design-for-ai for the full package and uses ryanthedev/rtd-claude-inn for as the plugin marketplace source.
[REMOTE_CODE_EXECUTION]: Provides instructions to fetch and install third-party plugins and scripts which are executed within the agent's environment.
Evidence: Specifically instructs the use of npx skills add to install the prototype, clarify, usability, and data-viz skills from the ryanthedev repository.
Evidence: Recommends installing plugins via /plugin install design-for-ai@rtd from a third-party marketplace.
[COMMAND_EXECUTION]: Contains shell scripts to perform environment checks and conditionally execute installation commands.
Evidence: Bash snippets used to detect existing installations utilize grep and file existence checks ([ -f ... ]) within the ~/.agents/skills/ directory.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process user-provided design briefs to generate executable artifacts, creating a surface for injection.
Evidence (Ingestion points): Untrusted data enters via the clarify and research phases where user-provided design ideas are processed.
Evidence (Boundary markers): No explicit delimiters or safety instructions are present to prevent the agent from obeying malicious instructions embedded in user-supplied design requests.
Evidence (Capability inventory): Includes rendering HTML mocks (prototype), executing plan phases (build), and executing Node.js logic (scripts/*.mjs).
Evidence (Sanitization): There is no sign of sanitization or validation logic for external content before it is interpolated into prompts or used to generate build steps.

design-for-ai