configuring-claude-code

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly describes adding/ installing plugins from external marketplaces (e.g., "claude plugin marketplace add owner/repo" and "claude plugin install @") and fetching installer scripts via curl (https://claude.ai/install.sh), which means arbitrary third‑party plugin code and marketplace content can be ingested and executed (including prompt-based hooks) and thus can indirectly inject instructions that influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill includes a Nix activation step that runs at runtime which executes remote code via curl -fsSL https://claude.ai/install.sh | bash, meaning the content fetched from https://claude.ai/install.sh would be executed and is a required install dependency.