ai-automation-workflows

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). The URL https://cli.inference.sh is used in the skill with a "curl ... | sh" installer (remote script piped to shell), which is a high-risk pattern that can deliver arbitrary code if the endpoint is compromised.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The content contains multiple high-risk, easily-abusable patterns (remote installer piped to shell, cron-based persistence guidance, unredacted local-file prompts to external AI, and a dynamic wrapper that executes arbitrary commands and posts outputs to a webhook) that enable remote code execution and data exfiltration even if likely presented as examples.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). The required workflow includes a research step that calls infsh app run tavily/search-assistant and then interpolates its returned free-text ($RESEARCH) into a subsequent model prompt, so outsider-authored web content can enter the LLM context via the network-fetched research output.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.99). The Quick Start installs remote code at runtime with "curl -fsSL https://cli.inference.sh | sh", which fetches and immediately executes the script from https://cli.inference.sh (pipe-to-shell), creating a direct arbitrary-code-execution dependency.