The Agent Skills Directory

[DYNAMIC_EXECUTION]: The build process in package-build.mjs and lib/preview-rebuild.mjs supports custom repository logic by dynamically importing JavaScript files from the project's .design-sync/overrides/ directory. This dynamic loading from project-controlled paths can lead to arbitrary code execution if malicious scripts are introduced into the synced repository.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of several Node.js packages (esbuild, ts-morph, playwright, typescript) and the download of the Chromium browser at runtime to facilitate the build and verification processes.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to manage dependencies, build Storybook assets, and run the internal toolchain. This includes npm install, npx storybook build, and the execution of the skill's own scripts via Node.js.
[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection. 1. Ingestion points: Untrusted data from repository documentation and source code enter the agent context through lib/docs.mjs and lib/source-kit.mjs. 2. Boundary markers: Explicit delimiters between code and documentation are absent in the generated prompt files, though the skill instructs the agent to ignore directives found in repository files. 3. Capability inventory: The skill uses spawnSync for system commands and possesses file writing and reading capabilities. 4. Sanitization: Ingested markdown content undergoes minimal transformation (frontmatter and HTML comment stripping) before being provided to the agent.

design-sync