The Agent Skills Directory

[NO_CODE]: The skill contains no executable scripts (such as Python, JavaScript, or Shell) or binary files. The entire package is composed of Markdown templates and YAML configurations for agent behavior.
[SAFE]: All external links provided in references/source-basis.md point to official domains of trusted academic organizations and publishers (nature.com, springernature.com, and datacite.org). These URLs are informational and do not involve code execution or hidden downloads.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user manuscript notes. \n
Ingestion points: User manuscript descriptions and Chinese author notes provided as input (SKILL.md, README.md).\n
Boundary markers: The skill defines specific boundaries for data, materials, and code availability.\n
Capability inventory: The skill does not use any tools or execution environments.\n
Sanitization: The instructions guide the agent to map user intent to professional templates rather than directly mirroring potentially malicious user input.

nature-data