The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references and fetches external media and summaries from goodcase.ai and x.com to ground the AI's design output in existing visual evidence.
Evidence: URLs in references/cases.md point to external case studies and media assets, such as https://goodcase.ai/cases/40-svg-5229d1feb246 and https://x.com/kawai_design/status/1991461556597715056.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and follows prompt excerpts from references/cases.md to generate its own artifacts, which represents an attack surface for instructions embedded in data.
Ingestion points: The references/cases.md file contains multiple prompt excerpts from various creators (e.g., @CryptoEights, @lumidotnew) that the agent is instructed to read and apply.
Boundary markers: The skill requires the creation of a 'Preserve / Replace / Avoid' reference contract before generating artifacts to explicitly separate case-derived traits from user-supplied context.
Capability inventory: The skill generates implementation-ready code and UI specifications (e.g., React, Tailwind, SVG generation).
Sanitization: Relies on a structured workflow and a verification checklist to ensure that the generated artifact aligns with user-defined success criteria and value propositions rather than being purely controlled by the external case data.

interactive-tool-simulation