The Agent Skills Directory

[SAFE]: No security issues were identified. The skill functions as a text-processing assistant for academic revision workflows. It does not request access to sensitive system files (e.g., .ssh, .aws), environment variables, or external network resources.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external reviewer comments and editor letters, which represents a surface for indirect prompt injection. However, the risk is negligible due to the absence of executable capabilities.\n
Ingestion points: Reviewer comments, editor decision letters, and author notes provided by the user are ingested into the agent context, as referenced in SKILL.md and references/intake-and-routing.md.\n
Boundary markers: The skill instructs the agent to preserve reviewer comments faithfully and assign stable identifiers (e.g., R1.1, R2.1), providing logical structure, though no explicit technical delimiters (like XML tags) are enforced for the input text.\n
Capability inventory: The skill's operations are strictly limited to text generation; it contains no subprocess calls, file system modification capabilities, or network exfiltration patterns across its instructions or referenced files.\n
Sanitization: No explicit input filtering or character escaping is implemented, relying on the platform's native safety guardrails.\n- [EXTERNAL_DOWNLOADS]: In references/source-basis.md, the skill provides informational links to official Nature and Springer Nature editorial guidelines. These are recognized as well-known academic services and the links are provided for guidance rather than for automated code downloads or runtime execution.

nature-response