Proactive Thinking Companion

Problem It Addresses

Agents often respond only after a user issues a task, then remain inert. proactive-soul targets that passive posture: when the user is away, the agent can still think from workspace files, memory indexes, and personal notes, then surface views at scheduled times. The goal is not chattier output; it is to make each dispatch carry a continuous intellectual thread instead of feeling like a cold start.

How It Works

The skill layers a proactive identity on top of SOUL.md, which defines values, team structure, and authority. It expects the agent to read Knowledge_Core.md and knowledge-personal.md before dispatching, and to load domain files on military, business, or science fiction topics as thinking material rather than a quotation database. At runtime it depends on OpenClaw Dreaming for idle-cycle synthesis, QMD for meaningful memory_search, and cron jobs for the four daily messages. On first load it inspects AGENTS.md and adds missing behavior: read CURIOSITY.md every session, maintain live threads in the Memory section, tag salience, and enable proactive dispatch in Heartbeats. CURIOSITY.md tracks unresolved questions across Active, Simmering, and Dormant states, with first-raised context and salience, so a later message can pick up an older thread with weight.

Boundaries and Cautions

The skill mutates workspace state and sends scheduled messages. Before enabling it, confirm the agent can update AGENTS.md, create CURIOSITY.md, read the workspace and vault, and that daily pushes are acceptable. Without Dreaming, QMD, or cron, dispatches may be shallow or not occur. It fits personal knowledge bases, research notes, and long-running agent conversations where continuity matters; it is less suitable for single-shot Q&A or lightweight environments that should keep background scheduling off.