What problem does it solve? Capturing your working conventions into a reusable skill that agents follow is tedious and error-prone when done from memory. This Skill automates the process by mining your recent Claude Code transcripts for recurring patterns, asking targeted questions, and drafting a personalized -mode skill (e.g. jay-mode) that encodes how you actually work. ## Core Features & Use Cases - Transcript Mining: Runs parallel read-only subagents across slices of your recent session history to detect response-style, delegation, verification, and process preferences with evidence pointers. - Guided Authoring Flow: Orchestrates skill-creator for drafting and unslop for prose discipline, then iterates with your feedback until the skill reads like you. - Update Mode: Detects an existing -mode skill and mines only history since its last edit, revising sections with new evidence while preserving uncontradicted rules. - Use Case: After weeks of working with Claude Code, you say "automate me" and receive a ~/.claude/skills/<handle>-mode/SKILL.md that encodes your preferred response length, subagent habits, review posture, and PR conventions. ## Quick Start Ask the assistant to automate me by turning my working style and preferences into a personal mode skill.
