Auto Explorer Self-Learning

Problem

When an AI meets an unfamiliar app, the hard part is not finding one tutorial, but lacking a stable learning loop: it may not know which apps exist on the machine, which are worth mastering, how to turn observed usage frequency into memory, or how to try operations safely. AI探索欲学习模块 turns AI learning into an executable workflow: on-demand learning, value-based depth, background usage watching, cross-platform environment discovery, and memory decay.

How It Works

The skill is organized around scripts/explorer.py, with these core steps:
- On-demand self-learning: when asked to use an app, the AI checks mastery, evaluates value, searches tutorials, and proceeds instead of refusing.
- Value-driven depth: learning_depth sets the target, while mastery_level records current ability, avoiding equal-depth study for every app.
- Environment discovery: Level 0 to Level 3 scan desktops, taskbars, start menus, PATH, user directories, registries, and more across Windows, macOS, and Linux.
- Usage watching: the Usage Watcher records low-signal facts such as process names to infer frequent apps, without screenshots, content access, or external upload.
- Memory decay: a three-level memory system keeps high-value knowledge and lowers stale entries over time.

Boundaries

The module requires Python 3.8+ and may use optional dependencies such as psutil, pywin32, pyautogui, and Pillow for enhanced detection or interaction. It enforces safety boundaries: it avoids System32, ProgramData, and other critical directories; asks before opening unknown apps or changing settings; and does not store passwords, keys, or personal file contents. Because background watching only records process names, it is not suited for deep content analysis, full UI visual recognition, or cross-machine knowledge migration; those remain planned capabilities.