agent-tools

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow ingests outsider-authored free text via the CLI-provided --input inline JSON or input file when running an app (e.g., prompts/queries/text), which is not constrained to a first-party/trusted source.

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

• Potentially malicious external URL detected (high risk: 1.00). The docs include an installer that pipes remote shell code into sh ("curl -fsSL https://cli.inference.sh | sh") and a manual install that downloads binaries/manifest from https://dist.inference.sh/cli/checksums.txt and https://dist.inference.sh/cli/manifest.json, which fetch and install remote executable code at setup time.