The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external content, specifically novel chapters and configuration files (e.g., 设定/characters.md, plan.md). This creates a surface where text written in a novel could attempt to influence the agent's behavior. However, the skill is constrained to providing a structured report and lacks dangerous capabilities such as network access or arbitrary command execution, which mitigates the risk.
[SAFE]: The skill does not contain any obfuscated code, hardcoded credentials, or external network requests. Its operations are limited to reading provided text files and generating a markdown report based on a predefined template.

novel-review