The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches Rails version metadata from rubygems.org to verify latest patch availability, which is a required pre-step in the upgrade workflow.
[COMMAND_EXECUTION]: Executes standard Rails development commands such as bundle exec, rails runner, rspec, and rake to verify application integrity and check for breaking changes.
[DATA_EXFILTRATION]: Transmits the content of the project's Gemfile.lock to api.railsbump.org (a service operated by the vendor, OmbuLabs) to generate gem compatibility reports.
[REMOTE_CODE_EXECUTION]: Downloads version metadata from rubygems.org and processes it via a ruby one-liner to extract the latest version number. This execution relies on data from a well-known service.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted application code and interpolates findings into prompts and reports, creating a surface for indirect prompt injection.
Ingestion points: Reads local files including Gemfile, Gemfile.lock, and the contents of the app/, config/, and lib/ directories (referenced in workflows/direct-detection-workflow.md).
Boundary markers: Employs markdown code blocks to delimit user-provided code within generated upgrade reports (referenced in templates/upgrade-report-template.md).
Capability inventory: Utilizes grep, glob, read, and executes shell commands like bundle exec and curl (referenced in SKILL.md and workflows/).
Sanitization: No explicit sanitization or filtering logic is specified for application content before it is included in report deliverables.

rails-upgrade