Proactive Claw

🦞 The most powerful proactive engine for OpenClaw. Your personal assistant that learns from you and helps you be more productive.

This skill appears to do what it says: a local-first proactive calendar assistant that reads your calendars and writes to a dedicated Actions calendar. Before installing, review these items: 1) credentials.json is required and contains your OAuth client secret — only provide a client you control and understand; token.json will store OAuth tokens locally, revoke them via your Google account if you stop using the skill. 2) The OAuth scope is full calendar access; the code claims to only write to an Actions calendar, but that restriction is enforced in code (so only use trusted copies of the skill). 3) setup.sh requires you to install Python packages manually from requirements-*.txt; consider installing into a virtualenv and prefer hash-locked dependencies. 4) Try the dry-run commands first (scripts/setup.sh --doctor, python3 scripts/action_planner.py --plan --dry-run, python3 scripts/action_executor.py --execute --dry-run, python3 scripts/quickstart.sh) to inspect planned actions before enabling the daemon. 5) If you care about maximum isolation, run the skill in a separate user account or VM, and audit config.json defaults (max_autonomy_level defaults to confirm). If you want additional assurance, request the full contents of the omitted files for manual review or run the audit commands listed in SECURITY.md.

Type: OpenClaw Skill Name: proactive-claw Version: 1.2.41 Proactive Claw is a comprehensive calendar management engine that demonstrates high-quality security engineering and transparency. Key safety features include a strict write-guard in `scripts/cal_backend.py` that restricts modifications to a specific 'Actions' calendar, local-first data persistence using SQLite (`memory.db` and `proactive_links.db`), and a setup process (`scripts/setup.sh`) that avoids risky patterns like `curl|bash` or automatic dependency installation. The bundle includes extensive documentation (`SECURITY.md`) detailing its threat model and data handling, and the code logic for autonomous actions is governed by user-configurable autonomy levels that default to 'confirm' mode.