critique-figures

Pass

Audited by Gen Agent Trust Hub on May 6, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's behavior is entirely consistent with its stated purpose of critiquing academic figures. It utilizes standard utilities and agent vision tools for legitimate analysis tasks.
  • [COMMAND_EXECUTION]: The skill instructs the agent to use the pdfimages command to check for raster components in PDF figures. This is a standard, low-privilege command used locally for the skill's primary function. The instruction also includes a fallback if the utility is not installed.
  • [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting and processing untrusted data from the user's project directory.
  • Ingestion points: Image files (PDF, PNG, SVG, etc.) and LaTeX manuscripts (SKILL.md) containing figure references.
  • Boundary markers: Absent.
  • Capability inventory: Shell execution (pdfimages) and vision tool usage (Read).
  • Sanitization: No explicit content-based sanitization or filtering is specified.
Audit Metadata
Risk Level
SAFE
Analyzed
May 6, 2026, 02:29 AM
Security Audit — agent-trust-hub — critique-figures