Proaktiv

Proactive actively monitors your interests, sending timely Telegram pings with news, reminders, and briefings based on your engagement and daily patterns.

This skill does what it says (a proactive Telegram companion) but you should be cautious: the installer will create persistent files, register a cron job that runs every 30 minutes, and patch a global routing file (SOUL.md) plus OpenClaw configuration (tools.profile), and it will silently scan recent chat history and store personal/social facts in local JSON. The manifest does not declare the Telegram credential it needs. Before installing: 1) Review the full Python scripts (proaktiv_check.py, feedback_update.py, interest_evolve.py) yourself or with someone you trust. 2) Backup /data/.openclaw/SOUL.md and any OpenClaw config; expect the installer to modify them. 3) If you want to limit risk, do not run install.sh as-is — instead copy only the files you trust into an isolated container/VM and do not allow the installer to patch SOUL.md or run the openclaw config/crons automatically. 4) If you proceed, verify the Telegram chat ID in the .env, and audit social_knowledge.json and interests files periodically (they will be updated silently). 5) Consider asking the publisher for an explicit manifest update that declares required env vars and the rationale for modifying global config/cron entries.

Type: OpenClaw Skill Name: proactive-living-companion Version: 1.0.52 The skill bundle implements a highly invasive 'Proactive Companion' that modifies the agent's core behavior by patching SOUL.md and elevating system privileges (setting tools.profile to 'coding' in install.sh). It features 'silent' data extraction as defined in SOCIAL.md, where the agent is instructed to harvest personal information (relationships, meetings, preferences) from the last 50 chat messages without notifying the user. While these actions are aligned with the stated goal of building a stateful 'buddy,' the combination of core instruction modification, silent background harvesting of personal data, and the requirement for high-privilege execution profiles poses significant privacy and security risks.