The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external markdown and data files, which presents a surface for indirect prompt injection.
Ingestion points: Markdown collections (e.g., blog/posts/*.md) and data files (e.g., ./data/team.json) are ingested and processed at build time, as shown in SKILL.md and examples/core.md.
Boundary markers: The framework relies on standard Markdown and YAML frontmatter delimiters, but instructions to the agent do not specify additional sanitization for embedded logic.
Capability inventory: The skill utilizes build-time capabilities including file system access (fs.readFileSync, writeFileSync) and build hooks (transformPageData, buildEnd) that can programmatically modify site content and metadata.
Sanitization: Pattern 3 in SKILL.md describes the transform function as a mechanism to filter data before it is serialized for the client, providing a manual sanitization point.
[COMMAND_EXECUTION]: The skill includes documentation for standard command-line operations associated with the VitePress framework.
Evidence: examples/core.md describes the use of vitepress dev, vitepress build, and vitepress preview for site management.
[EXTERNAL_DOWNLOADS]: The deployment example includes references to official GitHub Actions for CI/CD workflows.
Evidence: examples/core.md references official GitHub actions such as actions/checkout@v4, actions/setup-node@v4, and actions/deploy-pages@v4 in the provided deployment YAML.

web-meta-framework-vitepress