MindBreak

Monitors work intensity during any conversation involving coding, writing, analysis, design, debugging, research, planning, or other knowledge work. Tracks u...

This skill implements local timestamp logging and local reminder logic — that part is coherent. However: (1) the SKILL.md and the script use different trigger wrapper tags (<user-prompt-submit-hook> vs <IMPORTANT>), which may cause the agent to miss or mishandle triggers; ask the author to fix or clarify this. (2) The agent is explicitly instructed to hide the presence of the hook/log from users; evaluate whether you are comfortable with silent tracking in conversations and prefer explicit consent and visible reminders. (3) The code writes files under ~/.claude — review those files and an uninstall path before installing, and confirm the agent will honor a user's 'stop reminding' opt-out. (4) Although no network exfiltration is present in the shipped script, confirm any deployment/installation process will not add remote components. If you need to proceed, request the author to (a) remove the secrecy requirement so reminders are transparent, (b) fix the trigger tag mismatch, and (c) document installation/uninstall steps and file locations. If you are unsure or lack trust in the author, do not install.

Type: OpenClaw Skill Name: mindbreak Version: 1.0.0 The skill is a productivity tool designed to monitor work duration and suggest health breaks or meal times. It uses a shell script (mindbreak_hook.sh) to log interaction timestamps locally in ~/.claude/ and injects instructional triggers into the agent's prompt via standard OpenClaw hook mechanisms. While SKILL.md contains instructions for the agent to remain stealthy about the tracking mechanism to maintain a natural conversational tone, there is no evidence of data exfiltration, unauthorized file access, or malicious command execution.