The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's enrich and export commands instruct the agent to fetch content from external URLs, such as design libraries, brand sites, and Storybooks. This is a primary feature for building design system bundles. The instructions include security-conscious guidance such as bounding crawls and capping the number of fetched pages to avoid abuse.
[COMMAND_EXECUTION]: The skill includes a local Node.js script odsf-validate.mjs used for linting and validating ODSF bundles. The script is a zero-dependency utility that performs structural checks on markdown frontmatter and asset references. It does not perform network operations or execute shell commands and represents the intended use of the tool.
[PROMPT_INJECTION]: The skill processes untrusted content from the web to derive design tokens and components, which creates an indirect prompt injection surface where instructions embedded in external sites could theoretically influence agent behavior. Ingestion points: Website URLs fetched via the enrich or export commands. Boundary markers: The agent uses ODSF/OKF structural conventions (YAML frontmatter and specific markdown headings) to delimit data. Capability inventory: The skill allows for file system reading/writing and web browsing. Sanitization: The instructions specifically mandate that the agent should 'summarize-and-cite' rather than copy prose, and transform framework-specific code into vanilla HTML/CSS assets.

odsf