infographics
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The entry-point script
generate_infographic.pyusessubprocess.runto orchestrate an auxiliary internal Python script (generate_infographic_ai.py). Arguments are cleanly passed as structured lists rather than arbitrary shell strings, minimizing classic command injection vectors. - [EXTERNAL_DOWNLOADS]: The script uses the
requestslibrary to interact with OpenRouter endpoints (https://openrouter.ai/api/v1) for LLM/VLM querying and text-to-image/research generation. These are intentional operations matching the core scope of the skill. - [INDIRECT_PROMPT_INJECTION]: The skill ingests raw textual data provided by users or fetched from the web search/research module (
perplexity/sonar-pro) and dynamically inserts it directly into the next-step generation prompt strings insidegenerate_infographic_ai.py. This introduces a theoretical indirect prompt injection surface. However, since the output modality is purely image generation or structured review logs and boundary limits are maintained, the operational risk is limited.
Audit Metadata