The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to handle LLM tasks but contains no malicious internal instructions to bypass safety guidelines or override agent behavior. Input interpolation follows standard application patterns.\n- [DATA_EXFILTRATION]: Sensitive credentials (ANTHROPIC_API_KEY, OPENAI_API_KEY) are managed securely via environment variables. Network requests are constrained to well-known LLM provider domains and local services.\n- [EXTERNAL_DOWNLOADS]: Python requirements (anthropic, openai, httpx, tenacity) are reputable libraries from official repositories. No evidence of unverified remote code execution or suspicious script downloading was found.\n- [COMMAND_EXECUTION]: Static analysis shows no use of dangerous subprocess calls or shell execution. File system interactions are limited to safe configuration loading and skill structure validation using standard libraries.

llm-integration