The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process external design specifications (DESIGN.md) from a remote catalog and repository. These files contain natural language instructions that the agent is explicitly directed to follow as "hard constraints" during its UI construction process.
Ingestion points: Fetches an index file (llms.txt) and specific design entries (<slug>.md) from https://www.getdesign.kr and https://raw.githubusercontent.com/CaesiumY/ko-design-md/.
Boundary markers: The instructions lack explicit delimiters or "ignore embedded instructions" warnings for the fetched content, increasing the risk of the agent obeying instructions embedded within the design files.
Capability inventory: The skill uses curl for network and file operations and has the capability to write and modify the current project's codebase.
Sanitization: No validation or sanitization is performed on the fetched markdown data before it is interpreted by the agent.
[EXTERNAL_DOWNLOADS]: The skill retrieves design metadata, JSON tokens, and documentation from external sources.
Evidence: Uses curl to fetch content from https://www.getdesign.kr and https://raw.githubusercontent.com/CaesiumY/ko-design-md/.
Note: These resources trace back to the vendor/author's infrastructure (caesiumy) and are necessary for the skill's primary function.
[COMMAND_EXECUTION]: The skill provides exact shell command templates for the agent to execute for data retrieval and file management.
Evidence: The agent is instructed to use curl -fsSL --create-dirs -o ... with shell variables like $XDG_CACHE_HOME and to verify file sizes using wc -lc.

use-design-md