Self Improving.Bak

Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...

This skill appears to do what it says: keep a local, tiered memory and run conservative heartbeat maintenance. Before enabling it: (1) Back up any AGENTS.md, SOUL.md, HEARTBEAT.md or other workspace files it will edit, so non-destructive edits can't accidentally remove content. (2) Review ~/self-improving/ contents after initial setup to confirm nothing sensitive is being logged — the skill explicitly forbids storing credentials/financial/medical/third-party data, but you should verify. (3) Decline the optional 'Proactivity' install unless you trust the source, since that step will fetch/execute additional code. (4) If you use Windows, be prepared for some setup steps that assume POSIX commands; test in a controlled environment first. (5) If you want more certainty, ask the skill (or its author) to show a dry-run of file edits it will perform and to confirm exact commands it will run before making changes.

Type: OpenClaw Skill Name: self-improving-bak Version: 1.0.0 The skill implements a comprehensive self-improvement and memory system for OpenClaw agents, managing user preferences and corrections in a tiered storage structure within `~/self-improving/`. It includes detailed logic for memory promotion, demotion, and archiving, supported by a robust `boundaries.md` file that explicitly prohibits the storage of sensitive data such as credentials, financial info, or PII. While the setup process involves modifying core agent steering files (`SOUL.md`, `AGENTS.md`, `HEARTBEAT.md`) and using basic shell commands for file management, these actions are transparently documented and strictly aligned with the stated purpose of enhancing agent performance over time.