vault-to-artifact-tailor

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

• Third-party content exposure detected (high risk: 0.75). The required workflow ingests the audience constraint as raw text (category: “raw text of the constraint” provided by an outsider), and that text is then parsed by the agent at runtime into the LLM context for tailoring.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly accepts a user-provided GitHub repository URL as the "Vault locator" and uses the gh CLI at runtime to read vault markdown nodes whose content directly feeds and controls the agent's prompts and generation, so the user-supplied GitHub repo URL is a runtime external dependency that can control agent instructions.