The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources, specifically when 'reviewing a PR or finding for security issues', which creates a surface for indirect prompt injection attacks.
Ingestion points: The skill ingests content from Pull Requests and external security findings as described in the metadata.
Boundary markers: There are no instructions or boundary markers defined in the skill to isolate external data from the agent's primary instructions.
Capability inventory: The skill description implies the agent may perform sensitive actions such as 'managing secrets and API keys' and 'configuring CORS/CSP', which typically involve file system modifications or interactions with external APIs.
Sanitization: No specific sanitization or filtering instructions for external content are provided to the agent within this skill.
[SAFE]: The skill provides accurate and helpful documentation for e-commerce security, including the use of Spree's internal sanitization mechanisms (Spree::RichTextSanitizer) and the enforcement of Rails' encrypted credentials.
[SAFE]: References to default credentials for development environments are clearly labeled and accompanied by instructions to avoid their use in production settings, aligning with secure development practices.
[SAFE]: External references target trusted documentation sources such as official Ruby on Rails guides and vendor-specific documentation.

spree-security