The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate_infographic.py uses subprocess.run to execute a secondary AI-handling script. This is implemented safely using an argument list rather than a raw shell string, preventing shell injection vulnerabilities. The subprocess calls a local file within the skill's own directory.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to OpenRouter (openrouter.ai) to interact with established LLM models. These interactions are standard for the skill's purpose and target a well-known service provider.
[INDIRECT_PROMPT_INJECTION]: The skill includes an attack surface for indirect prompt injection because it ingests external data from web searches via Perplexity Sonar and interpolates it into prompts for image generation and quality review. However, the impact is confined to the content of the generated image and research logs, with no path to code execution or privilege escalation. The skill includes boundary markers to separate research data from instructions.
[CREDENTIALS_UNSAFE]: API keys are managed through environment variables (OPENROUTER_API_KEY). The skill includes clear instructions for users to set these variables in their local environment rather than hardcoding them, which follows security best practices.

infographics