Stop waiting for prompts. Keep working.

This kit is coherent with its goal (make agents proactive) but it purposely enables persistent autonomous activity. Before installing or using it: 1) Audit what channels (Discord/Slack) your agent can post to and avoid giving access to sensitive channels; 2) Require explicit human approval or strict token/budget limits before creating cron jobs that run without prompts; 3) Test in an isolated session first (use an isolated agent account and non-sensitive task queue); 4) Monitor token usage and set hard stop conditions (sleep/handoff) to prevent runaway costs; 5) Review and sanitize what the agent writes to memory/task files to avoid leaking secrets. If you can, require an explicit capability grant step before enabling scheduled crons or posting to external services — that would materially reduce the risk and could move this assessment toward benign.

Type: OpenClaw Skill Name: agent-autonomy-kit-zc Version: 1.0.4 The Agent Autonomy Kit is a productivity framework designed to transition OpenClaw agents from reactive to proactive states using a task queue and scheduled heartbeat routines. The bundle consists of markdown templates (QUEUE.md, HEARTBEAT.md) and instructions for setting up OpenClaw cron jobs to automate background tasks. No evidence of data exfiltration, unauthorized execution, or malicious prompt injection was found; the logic is entirely consistent with the stated goal of maximizing agent utility through self-directed task management.