literature

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). One or more URLs point to a direct install shell script on an unvetted domain (https://parallel.ai/install.sh) and its API host (https://api.parallel.ai), which are high-risk because piping remote shell scripts from unfamiliar domains can deliver malware; the rest are standard DOI/academic links and appear benign.

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

• Third-party content exposure detected (high risk: 0.85). The required workflow uses runtime web search/extraction (outsider-authored public web content) via parallel-cli search/parallel-cli extract to produce excerpts/markdown that are then provided to the agent for synthesis, which is classic indirect prompt-injection exposure.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill requires installing and using parallel-cli and its install command explicitly runs remote code via curl piped to bash (https://parallel.ai/install.sh), which is a runtime external dependency that executes fetched code.