obsidian-literature-workflow

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

• Third-party content exposure detected (high risk: 0.75). Yes—during the required canvas/graph build, the runtime script reads free-text from markdown files under the bound project’s Papers/*.md and Knowledge/*.md (e.g., path.read_text()), extracting fields like titles/concepts/methods/wikilinks/argument labels from those outsider-authored notes into the produced artifact that the agent context may later reference.