The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The skill utilizes npx shadcn@latest init (and framework-specific variants) in Step 3. This pattern downloads and executes code directly from the npm registry at runtime, which is an unverified remote execution vector.
[PROMPT_INJECTION] (HIGH): This skill exhibits a significant Indirect Prompt Injection surface (Category 8). It ingests untrusted data from Gemfile and package.json to determine its execution flow.
Ingestion points: Gemfile and package.json (SKILL.md Step 1).
Boundary markers: Absent; the skill does not use delimiters or instructions to ignore potential malicious content within the analyzed files.
Capability inventory: High. The skill can execute shell commands (bundle add, npx, rails generate) and write/modify files (CLAUDE.md, tsconfig.json, app/resources/application_resource.rb).
Sanitization: Absent; the skill relies on string matching within project files to decide which commands to execute without validating the content or source of those files.
[COMMAND_EXECUTION] (HIGH): The skill performs multiple arbitrary shell operations including bundle add [gems], rails generate, and npx initialization, which provide an attacker controlling the project files a path to execute arbitrary commands through the agent's environment.
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill promotes the installation of several external Ruby gems and Node packages. While these are common libraries, they are downloaded from external registries at runtime without version pinning or integrity checks in the instructions.

inertia-rails-setup