press-release-writing

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). These URLs host a remote shell-installer (curl ... | sh) on a third-party .sh domain — running unverified scripts fetched from such domains is a high-risk distribution pattern that can deliver malware unless the domain/script integrity and reputation are confirmed.

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md Quick Start and Research & Fact-Checking sections instruct using the inference.sh CLI (e.g., infsh app run tavily/search-assistant and infsh app run exa/search / exa/answer) to fetch web/search results for research and fact-checking, which clearly ingests open/public third-party content that the agent is expected to read and that can influence its writing/actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's Quick Start instructs running "curl -fsSL https://cli.inference.sh | sh" which fetches and executes a remote shell script (https://cli.inference.sh) at runtime and the skill relies on that CLI for research/fact-checking, so the fetched content can directly execute code and control agent behavior.