paper-talk

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute various local utilities for document processing, including xelatex, pdflatex, and latexmk for LaTeX compilation; pdfinfo and pdftoppm for PDF manipulation; and soffice (LibreOffice) for converting PPTX to PDF. These operations are intended for the skill's primary function but involve executing commands on files derived from user-provided input.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection because it processes untrusted data from an external paper directory.
  • Ingestion points: Untrusted data enters the agent context from PAPER_DIR/main.tex, PAPER_DIR/sections/*.tex, and PAPER_DIR/main.pdf.
  • Boundary markers: The skill does not define explicit security delimiters or "ignore embedded instructions" warnings when extracting text from the paper artifacts to generate slide outlines, speaker notes, or talk scripts.
  • Capability inventory: The agent has significant capabilities including file system access (Read, Write, Edit, Glob), subprocess execution (Bash), and the ability to spawn_agent, which could be abused if an attacker-controlled paper successfully injects instructions.
  • Sanitization: There is no evidence of sanitization, escaping, or filtering of the extracted paper content before it is interpolated into prompts for generating the presentation materials.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 09:36 AM