latex-posters

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill outlines a workflow to ingest external research content, such as scientific papers or outlines, and convert them into poster format, creating an indirect prompt injection attack surface.
  • Ingestion Points: Research papers, outlines, and abstracts provided by the user are ingested into the agent context during Stage 1 and Stage 4 of the workflow.
  • Boundary Markers: Absent. The templates and prompts do not specify structural boundaries or special delimiters to encapsulate untrusted user input.
  • Capability Inventory: The skill utilizes the Bash tool to execute LaTeX compilation utilities (pdflatex, lualatex, xelatex), Ghostscript (gs), and local helper scripts (scripts/review_poster.sh).
  • Sanitization: Absent. No filtering or escaping mechanisms are defined for user-supplied data before it is interpolated into prompts or LaTeX templates.
  • [COMMAND_EXECUTION]: The skill commands the execution of multiple command-line utilities via the Bash tool to support document compilation, automated quality checks, and package management.
  • Evidence: SKILL.md and scripts/review_poster.sh describe invocations of pdflatex, pdfinfo, pdffonts, pdfimages, gs (Ghostscript), and tlmgr install for package setup. These operations utilize standard local environment binaries required for LaTeX building workflows.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 02:39 PM