The Agent Skills Directory

[SAFE]: The skill provides structured guidelines and templates for agent development without introducing security risks. No malicious patterns, obfuscation, or instructions to bypass safety guidelines were detected.- [COMMAND_EXECUTION]: The instructions include standard platform management commands (such as 'openclaw gateway restart') and the execution of utility scripts located within the platform's global installation path. These are expected behaviors for a development-focused skill.- [EXTERNAL_DOWNLOADS]: An example skill ('weather-bot') utilizes 'curl' to interact with 'wttr.in', which is a well-known and documented service for providing weather data.- [DATA_EXFILTRATION]: No unauthorized transmission of sensitive data was identified. The 'Auto Handoff' system and context recovery mechanisms use internal platform tools (e.g., 'sessions_send', workspace file writes) to maintain legitimate conversation history and memory.

agent-forge