gemini3pro

[Skill Scanner] Pipe-to-shell or eval pattern detected All findings: [CRITICAL] command_injection: Pipe-to-shell or eval pattern detected (CI013) [AITech 9.1.4] [CRITICAL] command_injection: URL pointing to executable file detected (CI010) [AITech 9.1.4] [HIGH] command_injection: Reference to external script with install/setup context (SC005) Functionally, this skill is a client that forwards user prompts and optional images to a third-party API (grsai.com) to obtain text or vision responses. No direct evidence of obfuscated code or explicit malware is present in the provided manifest. The primary supply-chain/privacy risk is that all user data and credentials are routed through an undocumented third-party gateway; this is potentially dangerous if the gateway operator is untrusted or compromised. Recommend: do not use with sensitive prompts, images, or production API keys until the grsai.com operator and privacy practices are verified; inspect the actual script source for request handling, logging, and persistence behaviors; prefer calling official Google/Google Cloud Gemini endpoints if direct use of Google services is intended. LLM verification: The provided SKILL.md is a lightweight integration document that, on its face, does not contain in‑file malicious code; however it documents sending sensitive inputs (prompts, images, API key) to a third‑party gateway (grsai.com) while recommending executing an unsigned remote installer (astral.sh) via 'curl | sh'. Those two facts constitute significant supply‑chain and data‑exposure risks. Treat this package as untrusted for sensitive data until the actual client implementation and installer sc