paper-figure

Warn

Audited by Gen Agent Trust Hub on Jul 18, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill workflow involves generating standalone Python scripts based on content parsed from external files (PAPER_PLAN.md and experiment data) and then executing them. This pattern allows for arbitrary code execution if the source files are maliciously crafted.- [COMMAND_EXECUTION]: The skill uses a bash loop to execute all generated scripts (gen_fig*.py) in the local environment, which is a dangerous execution pattern that lacks granular control or safety checks.- [PROMPT_INJECTION]: The skill ingests data from untrusted files and interpolates it directly into a prompt for a review model via the Codex MCP tool. This occurs without sanitization, delimiters, or boundary markers, creating a surface for indirect prompt injection.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 18, 2026, 04:51 PM
Security Audit — agent-trust-hub — paper-figure