agent-tools

[Skill Scanner] URL pointing to executable file detected All findings: [CRITICAL] command_injection: URL pointing to executable file detected (CI010) [AITech 9.1.4] [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] No direct malicious code is present in the provided skill document. The content documents a legitimate-seeming CLI tool (inference.sh) that performs remote model/app execution. Primary risks stem from the recommended curl|sh install pattern, broad capabilities (including social posting) that require credentials, and ambiguity about whether the CLI proxies requests or transmits credentials to third parties. Recommend: inspect the installer script before running, prefer the manual install with checksum/sig verification, review how infsh stores/transmits credentials, and audit the downloaded binary and network endpoints before trusting automation that can post or act on behalf of a user. LLM verification: This skill documentation describes a CLI-based platform (inference.sh) that legitimately runs many cloud AI apps and includes potentially risky installation and data flows. The main supply-chain concerns are (1) the recommended curl | sh installer which executes remote code without mandatory verification, and (2) centralization of prompts, uploaded media, and third-party API actions through inference.sh/cloud.inference.sh which could expose data or credentials depending on how authentication is