The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references several external URLs for academic guidelines (e.g., nature.com and springernature.com). These are well-known, trusted academic publishing services. The skill does not perform automated downloads or execute remote scripts; the URLs are provided as reference documentation for the agent's logic.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data such as reviewer comments and editor letters. While this represents a potential injection surface, the skill includes extensive 'Red lines' and a 'QA checklist' that instruct the agent to remain faithful to reviewer text, avoid fabrication, and maintain a professional tone. Furthermore, the skill lacks access to sensitive tools (such as shell execution or network requests), mitigating the impact of any potential injection.
[NO_CODE]: The skill consists entirely of Markdown instructions, reference documents, and test fixtures. It does not include any Python, Node.js, or shell scripts, eliminating risks associated with dynamic code execution or privilege escalation.
[DATA_EXPOSURE_AND_EXFILTRATION]: No patterns for accessing sensitive local files (such as SSH keys or environment variables) or hardcoded credentials were found. The skill does not utilize any network-capable tools that could be used for data exfiltration.

nature-response