powerpoint-automation

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches and extracts content from public web articles/blogs (e.g., "From Web Article" example in SKILL.md and the fetch/extract steps in references/instructions/purpose-blog.instructions.md and instructions/common.instructions.md), producing content.json that is then summarized, translated, and used to drive BUILD/creation actions—so untrusted, user-generated web content is ingested and can materially influence subsequent tool use and decisions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly fetches web article content at runtime (e.g., https://zenn.dev/example/articles/azure-copilot-guide) using curl/fetch_webpage/microsoft_docs_fetch to build content.json that is injected into agent context and directly controls model output.