The Agent Skills Directory

[SAFE]: The skill explicitly implements a 'User-facing privacy rule' in SKILL.md and references/figure-contract.md, instructing the agent to never disclose private local paths, filenames, or internal template identifiers in its output.
[SAFE]: External references are limited to official Nature research guidelines and reputable academic repositories. These are documented neutrally and do not involve untrusted remote code execution.
[SAFE]: The skill's architecture includes a 'Backend selection' blocking gate that prevents the agent from executing code in an unvetted or unintended runtime, ensuring it sticks to the user's preferred environment (Python or R).
[SAFE]: The provided code patterns in api.md and tutorials.md utilize standard, industry-recognized libraries (e.g., matplotlib, seaborn, ggplot2, patchwork) without any signs of obfuscation, credential harvesting, or suspicious dynamic execution.

nature-figure