The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's subagents (SimpleCov and RubyCritic) temporarily modify the project's Gemfile to include these tools and run bundle install. These are well-known, industry-standard Ruby gems fetched from the official RubyGems registry.
[COMMAND_EXECUTION]: The skill executes several shell commands to facilitate the audit, including bundle exec rspec, bundle exec rails test, bundle exec rubycritic, and file management commands like cp, rm, and git stash. These commands are integral to the skill's primary function of auditing code and are executed only after the user opts in through a confirmation prompt.
[PROMPT_INJECTION]: The skill identifies an indirect prompt injection surface as it ingests and processes untrusted source code from the target application during the audit process.
Ingestion points: Source files and configuration files throughout the project directories (app/, config/, db/, lib/, spec/, test/).
Boundary markers: No explicit delimiters or instructions to ignore embedded content are used during code analysis.
Capability inventory: The skill can execute shell commands (bundle exec), modify local files, and read project structure.
Sanitization: There is no evidence of specific sanitization or escaping applied to the ingested code snippets before they are analyzed by the LLM.

rails-audit