paper-poster

Warn

Audited by Gen Agent Trust Hub on Jul 18, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill generates a local Python script (poster/generate_pptx.py) from provided templates and executes it at runtime to handle PowerPoint generation.
  • [DATA_EXFILTRATION]: The instructions direct the agent to check for the existence of ~/.claude/feishu.json to send optional notifications. Accessing credential or configuration files in hidden platform-specific directories constitutes a data exposure risk.
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions to download the TeX Live installer from CTAN (mirror.ctan.org) and execute it directly via the shell to set up the LaTeX environment. While CTAN is a well-known service, the pattern of downloading and executing remote installers is a high-privilege operation.
  • [EXTERNAL_DOWNLOADS]: The skill installs several third-party Python packages (pdf2image, python-pptx, pymupdf) and LaTeX packages (tcolorbox, tikzfill, etc.) from external registries during the setup phase.
  • [PROMPT_INJECTION]: The skill processes untrusted content from the user's paper (paper/sections/*.tex) and sends it to a model for review without explicit sanitization or boundary markers, creating a surface for indirect prompt injection where instructions embedded in the paper could influence the agent's behavior.
  • [COMMAND_EXECUTION]: Extensive use of shell commands through the Bash tool to manipulate files, manage package installations, and compile LaTeX documents using latexmk and pdflatex.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 18, 2026, 04:51 PM
Security Audit — agent-trust-hub — paper-poster