The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local project files like Gemfile and package.json to determine which features to enable and which dependencies to recommend. While these are standard project files, they act as external ingestion points that influence the execution of shell commands and file modifications.
Ingestion points: Reads Gemfile and package.json in Step 1 to detect the existing stack; reads CLAUDE.md in Step 4 to identify existing configuration blocks.
Capability inventory: The skill can execute shell commands (bundle add, npx shadcn init) and perform file writes to various locations (app/resources/application_resource.rb, config/initializers/typelizer.rb, tsconfig.json, CLAUDE.md).
Boundary markers: None are specified for the parsing of the manifest files.
Sanitization: No specific sanitization logic is provided for the contents of the manifest files beyond simple string matching for expected package names.
[COMMAND_EXECUTION]: The skill performs shell command execution to install new dependencies and initialize UI frameworks. These actions are standard for a setup utility and are mitigated by an interactive prompt (AskUserQuestion) that requires user approval before the commands are executed.
Evidence: Uses bundle add for Ruby gems (alba, typelizer, alba-inertia, js-routes, pagy) and npx shadcn@latest init (including Vue and Svelte variants) for frontend components.

inertia-rails-setup