paper-poster

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

• Third-party content exposure detected (high risk: 0.75). Yes: the required workflow reads outsider-authored free text from paper/sections/*.tex (compiled paper sections provided by the user’s paper, not authored by the operating user) during “Phase 1: Content Extraction,” and then feeds the extracted content into LLM steps (e.g., generating POSTER_CONTENT_PLAN.md and later Codex MCP review prompts that include [PASTE POSTER_CONTENT_PLAN.md] and [PASTE key sections of main.tex]).

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

• Potentially malicious external URL detected (high risk: 1.00). The skill includes a runtime installation step that downloads and extracts a remote installer which is then executed (curl -L https://mirror.ctan.org/systems/texlive/tlnet/install-tl-unx.tar.gz | tar xz followed by ./install-tl), so the URL https://mirror.ctan.org/systems/texlive/tlnet/install-tl-unx.tar.gz is fetched and run during skill runtime.