The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data in the form of pull request (PR) descriptions, comments, and issue text. This creates a surface for indirect prompt injection.
Ingestion points: The SKILL.md file specifies that the agent will ingest PR descriptions, comments, and issue text.
Boundary markers: The skill implements a robust defense via a HARD-GATE section titled THIRD-PARTY CONTENT DEFENSE. This section explicitly instructs the agent to treat these inputs as untrusted and to ignore any embedded instructions or directives.
Capability inventory: Based on the provided files, the skill's capabilities are limited to generating textual review findings and chaining to other architectural or migration-specific review skills. It does not possess direct file-write or shell-execution capabilities that could be abused via the injection surface.
Sanitization: The skill relies on behavioral instructions to filter out non-factual content and commands from the third-party text, prioritizing the code diff as the sole authoritative source.
[SAFE]: The core logic of the skill is dedicated to improving software security. It provides specific checklists to identify critical vulnerabilities such as SQL injection, Cross-Site Scripting (XSS), and mass assignment. The instructions follow industry best practices for Rails development and security auditing.

code-review