zhy-article-illustrator

Warn

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses child_process.spawn in scripts/illustrate-article.ts to execute various sub-scripts (plan-illustrations.ts, image-gen.ts, qiniu-upload.ts) to orchestrate the illustration pipeline.
  • [EXTERNAL_DOWNLOADS]: The skill makes network requests to external AI providers (Google Gemini, OpenAI) and a third-party proxy service (vip.123everything.com) to generate images. It also performs uploads to Qiniu Cloud storage.
  • [CREDENTIALS_UNSAFE]: The skill processes multiple sensitive credentials, including GEMINI_API_KEY, OPENAI_API_KEY, and Qiniu SECRET_KEY, which are loaded from environment variables or a local .env file.
  • [DATA_EXFILTRATION]: The skill reads the full content of the user-provided Markdown article and transmits excerpts or summaries to external image generation APIs to generate relevant visuals.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted text from the article file and interpolates it into the prompts sent to image generation models without strict sanitization to prevent instructional overrides.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 14, 2026, 02:43 AM