The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute development commands including bundle exec rspec and bundle exec rubocop to test and lint the generated code. These are standard operations for a Ruby on Rails development assistant.
[EXTERNAL_DOWNLOADS]: The UserPresenter and PostPresenter code templates include references to well-known external services. Specifically, it generates URLs for Gravatar (gravatar.com) for avatars and Twitter (twitter.com) for social sharing. These are neutral, well-known service integrations.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it is designed to ingest and format data from ActiveRecord models (e.g., Post#body, User#email) that may contain untrusted user content.
Ingestion points: Data is read from model attributes in app/presenters/ files.
Boundary markers: The generated presenter code does not include explicit instructions to ignore embedded prompts within the formatted data.
Capability inventory: The agent can create/modify files in app/presenters/ and spec/presenters/, and execute shell commands via bundle exec.
Sanitization: The code leverages standard Rails helpers like content_tag, link_to, and CGI.escape which provide HTML escaping and URL encoding, reducing the risk of XSS or malformed URL injection.

presenter-agent