pandoc

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

• Third-party content exposure detected (high risk: 0.70). The skill explicitly instructs fetching and using external, community-hosted resources (e.g., "curl -O https://raw.githubusercontent.com/citation-style-language/styles/master/apa.csl") and allows Pandoc to load external templates, CSL/bibliography files and filters (--csl, --bibliography, --lua-filter), which the agent will read and apply at runtime, exposing it to untrusted third‑party content.