modules

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

• Third-party content exposure detected (high risk: 0.75). High risk: the workflow fetches outsider-authored free text from public sources at runtime—module README/snippets from https://registry.coder.com/modules/... (or a local clone of coder/registry) and potentially template/module search results—then inserts that readable Terraform text into the agent’s LLM context for processing.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly requires runtime fetching/searching of module content from the Coder registry (e.g., https://registry.coder.com/mcp and README pages such as https://registry.coder.com/modules//README.md) and uses that fetched canonical snippet to construct/insert Terraform module code, meaning remote content directly controls the agent's output.