The Agent Skills Directory

[Prompt Injection] (SAFE): No patterns of instruction override, bypass, or adversarial role-play were detected. The instructions are focused on providing a professional evaluation persona.
[Data Exposure & Exfiltration] (SAFE): There are no hardcoded credentials, sensitive file path accesses, or network-enabled tools. The skill only utilizes the 'Read' tool to access its own reference documentation and user input.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill does not install any external packages or attempt to execute remote scripts. No dynamic code generation (eval/exec) is present.
[Indirect Prompt Injection] (LOW): The skill ingests untrusted story text from users. While this is an entry point for indirect injection, the skill lacks the capability to perform dangerous actions (like file writing or network requests) based on that input, and the instructions include constraints to keep the agent focused on evaluation only.

novel-evaluator