The Agent Skills Directory

[SAFE]: The skill manifest and instructions reference only local markdown documents within the skill's own directory structure. There are no external network requests, remote script downloads, or untrusted package dependencies.- [SAFE]: Comprehensive review of the skill's logic reveals no evidence of malicious prompt injection, base64 obfuscation, or hidden instructions intended to bypass safety guardrails.- [SAFE]: The skill does not define or invoke any privileged tools, shell commands, or filesystem operations, effectively eliminating the risk of command injection or unauthorized data access.- [SAFE]: All instructional content is focused on academic assessment, with explicit mandates to ground findings in user-provided evidence and avoid the invention of reviewer identities or data.

nature-reviewer