inno-idea-eval

SUSPICIOUS. The skill's core behavior fits its stated purpose, but it relies on an unverifiable transitive dependency in ~/.claude/skills/searching-ai-papers and ingests untrusted external content into a write-capable evaluation pipeline. No credential harvesting, exfiltration endpoint, or covert real-world action is evident, so this is better classified as a high-risk workflow dependency/prompt-surface issue than malware.