The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches documentation, graph templates, and installation manifests from https://mdx-graphs.kshv.me at runtime to guide its output.
[COMMAND_EXECUTION]: Provides instructions to execute shell commands for package management, including pnpm add knap and the use of pnpm dlx for running the shadcn CLI.
[REMOTE_CODE_EXECUTION]: Recommends the use of shadcn add targeting a remote JSON manifest at https://mdx-graphs.kshv.me/r/all.json, which downloads and integrates external component code into the user's project.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from an external llms.txt file to determine its output format and content. * Ingestion points: https://mdx-graphs.kshv.me/llms.txt used in SKILL.md and recipes.md. * Boundary markers: None. The agent is instructed to directly copy and swap labels from the external content. * Capability inventory: Access to shell execution for package installation (pnpm) and component management (shadcn). * Sanitization: No validation or escaping of the remote content is performed before interpolation into the agent's context.

markdown-graphs