The Agent Skills Directory

[DYNAMIC_EXECUTION]: The docs codeblocks validate command enables the execution of code snippets extracted from documentation (MD/MDX files) using various runners such as local environments, Vercel sandboxes, E2B, or Daytona. This is an intended primary feature for documentation quality assurance.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documentation content which could theoretically contain malicious instructions or code snippets intended for the validation runner.
Ingestion points: Reads and parses Markdown/MDX files within the user's project directory.
Boundary markers: Uses code fence metadata (e.g., runnable) to identify candidate blocks for execution.
Capability inventory: Includes file system writes (scaffolding), network communication (cloud/search sync), and arbitrary code execution via configured runners.
Sanitization: The instructions recommend using the --plan flag before execution to review planned actions and require explicit configuration for runner providers.
[COMMAND_EXECUTION]: The skill relies on standard package manager runners (npx, pnpm dlx, yarn dlx, bunx) to invoke the latest version of the vendor's CLI tool for various lifecycle tasks.
[EXTERNAL_DOWNLOADS]: The CLI interacts with several well-known services and platforms for its operational needs, including Vercel, OpenAI, Typesense, Algolia, and E2B for deployment, indexing, and code execution environments.

cli