external-llm-consulting

SUSPICIOUS. The overall purpose is coherent: consulting external LLMs via their CLIs fits the stated goal, and the skill explicitly warns against sending secrets. Codex usage aligns with official OpenAI distribution and expected data flow. However, the Gemini install instruction is internally inconsistent and points to a package scope that does not match the cited Google Gemini publisher/docs, creating a significant install-trust and credential/data-forwarding concern. Because the skill causes project context to be sent to external vendors and one CLI provenance is mismatched, it should not be treated as fully benign.