Self-Direction

Your agent learns to think like you. Captures your direction system, makes decisions as you would, guides all processes toward your goals.

This skill is internally coherent with its purpose, but it stores persistent logs and a model of how you make decisions in ~/self-direction/, which can include sensitive context. Before installing or enabling it, consider: (1) Do you trust the agent/process that will read and update those files? (2) Review the ~/self-direction/ files regularly and delete or redact anything sensitive. (3) If needed, restrict file permissions or store the directory on encrypted storage. (4) Limit what the agent is allowed to observe (avoid automatic logging of private chats, credentials, or financial data). (5) If you want tighter control, require explicit approval for high-stakes decisions and configure clear escalation triggers in the transmission frames. If any of these are unacceptable or unclear, treat the skill as privacy-sensitive and review or decline installation.

Type: OpenClaw Skill Name: self-direction Version: 1.0.0 The skill bundle is classified as benign. All files describe an AI agent's internal learning process for understanding user preferences and making autonomous decisions, with clear instructions for user interaction and data management within the designated `~/self-direction/` directory. There is no evidence of data exfiltration, malicious execution, persistence mechanisms, or prompt injection designed to subvert the agent's core function. The `_meta.json` explicitly states no external binaries are required, and the markdown instructions consistently reinforce a collaborative, user-validated learning loop, with `clawhub install` commands requiring user confirmation.