The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a CLI wrapper to execute its core AI logic in a secondary script via subprocess. This is implemented securely using argument lists and a restricted set of forwarded environment variables to ensure credentials from the parent environment are not leaked to the child process.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from user prompts and external research results gathered via Perplexity Sonar. It mitigates potential prompt injection by employing clear boundary markers and headers (e.g., 'USER REQUEST', 'RESEARCH CANDIDATES') to separate instructions from data.\n- [DATA_EXPOSURE]: The skill includes logic to resolve the OpenRouter API key from environment variables or local .env files. To prevent accidental leakage, the implementation specifically redacts this sensitive key from all error messages and debug logs.\n- [EXTERNAL_DOWNLOADS]: The skill communicates with well-known services (OpenRouter and Perplexity) to perform research and generate images. These external operations are essential for the skill's intended purpose and target established, trusted endpoints.

infographics