ai-rag-pipeline

[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: Natural language instruction to download and install from URL detected (CI009) [AITech 9.1.4] [CRITICAL] command_injection: Natural language instruction to download and install from URL detected (CI009) [AITech 9.1.4] This skill is functionally coherent and aligned with its stated purpose: it composes search/extraction apps with LLMs via infsh to build RAG pipelines. There is no direct evidence of obfuscated or explicitly malicious code in the provided content. The primary security concerns are: (1) the install-by-curl|sh pattern (requires users to trust the remote script and distribution host), (2) routing of all prompts, search results, and any supplied credentials through the inference.sh platform and hosted apps (a privacy/credential risk if users expect direct API calls), and (3) the broad allowed-tools permission (Bash(infsh *)) which grants wide execution scope. Recommend users inspect the installer, review inference.sh's data handling and storage policies, and restrict permission scope where possible. Overall: benign functionality but moderate operational risk due to data routing and installer/execution model. LLM verification: This skill documentation and examples are coherent with the stated purpose (building RAG pipelines using hosted search and LLM apps). The primary security concerns are supply-chain and privacy/trust-related: the instruction to run a remote installer via curl | sh and the fact that example pipelines route queries and retrieved content through inference.sh/infsh and hosted app backends. Those patterns are not direct evidence of malware, but they are high-risk from a supply-chain and data-exposure