The Agent Skills Directory

[DATA_EXFILTRATION] (SAFE): The scripts operate exclusively on the local file system for reading skill contents and writing ZIP archives. No modules for network communication (like requests or urllib) are imported or used.\n- [COMMAND_EXECUTION] (SAFE): The analysis found no instances of subprocess spawning, shell commands, or os.system calls. All logic is implemented using pure Python functions.\n- [REMOTE_CODE_EXECUTION] (SAFE): No dynamic execution sinks such as eval() or exec() are present, and the skill does not attempt to download or run remote scripts.\n- [CREDENTIALS_UNSAFE] (SAFE): No hardcoded credentials, API keys, or access to sensitive environment/SSH files were detected in the source code.\n- [PROMPT_INJECTION] (SAFE): The scripts are utility tools and do not interact with agent prompts or process untrusted user input intended for an LLM.

skill-creator