content-repurposing

[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] No clear malicious code or backdoor is present in this skill file. The content and example commands are coherent with the stated purpose (content repurposing) and the capabilities demonstrated. The primary security concerns are: 1) use of curl | sh installer (supply-chain risk — requires trusting inference.sh/dist.inference.sh and their checksum hosting), 2) centralized routing of content and credentials through the inference.sh platform (privacy/credential trust risk), and 3) broad allowed-tool permissions for the infsh CLI. Recommend: verify the installer checksum from an independent source before running, review infsh login/scopes to minimize credential exposure, and prefer explicit app endpoints or self-hosted alternatives if handling sensitive content. LLM verification: No direct malicious payloads or obfuscated code were found in the skill text. The primary supply-chain risk comes from the installer pattern (curl | sh) and the design choice to route all model inference and social publishing through the inference.sh managed service. That centralization means trusting inference.sh/dist.inference.sh with uploaded content and credentials. This makes the skill SUSPICIOUS from a supply-chain perspective — acceptable if the user trusts and independently verifies the