coderabbit

[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) This skill/document is functionally coherent for using a remote AI code-review CLI, but it contains supply-chain and data-exfiltration risks: it instructs executing a remote installer via curl|sh without integrity checks, and it relies on a remote service for reviews (local repo content will be sent over the network). No direct evidence of malware or obfuscation exists, but the install-and-send pattern is risky for sensitive code. Recommend: avoid running curl|sh without inspection, require installer signing/checksums, warn users explicitly that repository code will be transmitted, avoid writing sensitive outputs to world-readable /tmp, and provide an option for fully local/offline review. LLM verification: The skill/instructions are coherent and implement a convenient workflow for running CodeRabbit CLI reviews, but they use high-risk supply-chain and privacy patterns: piping an install script from a remote host to sh, lack of installer verification, unclear dataflow (likely uploading repository contents), persistent authentication without storage/TTL details, and writing unredacted outputs to /tmp. I do not find explicit malicious code in the fragment, but the instructions present moderate securi