The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill automatically clones its runtime component from the author's GitHub repository (theclaymethod/artifacture) to a local directory (~/.artifacture) if it is not already present, followed by an npm install command to resolve dependencies.
[COMMAND_EXECUTION]: To process source code and generate artifacts, the skill executes various local tools and scripts, including git, node, npm, hyperframes, and ffmpeg. It also supports running project-defined build and export scripts from package.json during the artifact generation process.
[DYNAMIC_EXECUTION]: A local preview server (scripts/preview.mjs) is provided to enable real-time visualization of artifacts. This server includes functionality to write text changes directly back to the project's source files (HTML, MDX, TSX) when the user performs edits within the preview interface. The server implements security measures such as loopback-only binding, session-based authorization via cookies, and same-origin validation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the codebase, such as git diffs and plan documents. This attack surface is mitigated through specific instructions to the agent's sub-components to prioritize sanitization (HTML escaping) and to ignore any instructional content embedded within the source data.

visual-explainer