auto-extract

What problem does it solve?

This Skill reduces repeated trial-and-error by automatically analyzing your current or historical AI coding sessions and extracting high-signal learning to persist as configuration knowledge.

Core Features & Use Cases

• Dual-mode learning: Uses the session data by default, or switches to batch analysis when the input explicitly starts with "历史分析".
• Signal-driven extraction: Identifies progress wins, failure points, explicit user corrections, and inferred preferences from messages across sessions.
• Four-layer persistence model: Routes each learning item into the right storage layer (L1 user configuration, L2 project root, L3 submodule, or L4 situation memory) with cross-layer de-duplication and user confirmation before writing.
• Robust environment probing: Detects available agent configuration files and memory backends without hardcoding paths, then creates or updates the most semantically appropriate targets.

Quick Start

Ask your AI to run auto-extract for the current conversation so it can extract learning signals and propose where to persist them, then confirm the suggested writes.