baoyu-article-illustrator
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted article content provided by users (Step 2: Analyze content). This poses a theoretical risk of indirect prompt injection where an article could contain instructions designed to influence the image generation prompts. However, the skill mitigates this by using structured prompt templates (ZONES, LABELS, COLORS) and strictly limiting the generation output to raster images, preventing any code-based feedback loop.
- [COMMAND_EXECUTION]: The skill contains logic to execute local wrapper scripts (references/codex-imagegen.md) using 'bun' or 'npx'. This execution is part of the core functionality for interfacing with the Codex image generation CLI and follows established vendor patterns for the 'jimliu' ecosystem, involving dynamic discovery of local assets rather than downloading untrusted remote code.
Audit Metadata