The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data in the form of existing application codebases, creating a potential surface for indirect prompt injection.
Ingestion points: The skill utilizes Read, Grep, and Glob tools to scan project files, including controllers, models, and services, as part of its core analysis and implementation workflow (as seen in SKILL.md and references/analyze.md).
Boundary markers: The instructions lack specific boundary markers or delimiters to help the agent distinguish its internal instructions from the content of the files being analyzed.
Capability inventory: The agent has the capability to read extensive parts of the file system and modify the codebase using the Task tool based on the patterns it detects.
Sanitization: No sanitization, escaping, or filtering of the ingested code content is described in the skill's logic.
[SAFE]: The skill uses well-known external resources for documentation and reference, such as the official Ruby on Rails guides and the Basecamp/37signals GitHub repository for the 'Fizzy' project. These references are documented neutrally and originate from trusted service providers.

vanilla-rails