v1.0.0 — Rewrite & Apply + Proactive Scaffold

New features:
• "Rewrite & Apply" button — rewrites your prompt and replaces the text directly in the input box. No copy-paste needed.
• Task-aware scaffold — detects if you're debugging, implementing, refactoring, testing, or reviewing, and adds the specific structural slots your prompt is missing (e.g., Error/File/Expected for debug prompts).
• Proactive flyout — when your prompt scores below BASIC, suggestions auto-appear while you type (once per turn, non-intrusive).
• Specificity scoring — detects I/O specifications and edge case mentions, backed by research (Zi+ 2508.03678).
• Improved flyout behavior — click badge to keep open, proactive popup auto-hides after 6s, Apply closes immediately.

Scoring engine now calibrated against 10 peer-reviewed research papers.

v0.6.1

• Click the score badge for inline improvement suggestions with point values
• Popup now shows live draft score while typing (before submitting)
• Fixed: badge not appearing on ChatGPT/Claude.ai (React contenteditable)
• Fixed: scoring accuracy aligned with CLI (position detection)
• Smart filtering: conversational turns (yes, ok, thanks, 继续, はい) are skipped — only real prompts get scored

v0.6.0 — Inline Suggestions

• Click the live score badge to see improvement tips with point values
• Top 3 suggestions shown in a flyout panel (auto-hides after 8 seconds)
• Suggestions update in real-time as you type
• All analysis runs locally — no data sent anywhere

Previous (v0.5.x):
• Score badge shows tier labels (EXPERT/STRONG/GOOD/BASIC/DRAFT) instead of raw numbers
• Rebalanced scoring: plain-text prompts score fairly without needing markdown

v0.5.3 — Tier Labels + Popup Update

• Popup now shows tier labels (DRAFT / BASIC / GOOD / STRONG / EXPERT) everywhere instead of raw numbers
• Score dimensions reordered: Clarity first (most impactful)
• Dimension max values updated to match rebalanced scoring weights
• Color thresholds adjusted for fairer grading

Previous features: Live scoring badge, prompt capture, 5-dimension analysis, privacy detection, compression, CLI sync.

v0.5.2 — Scoring Rebalance + Tier Labels

NEW: Live scoring badge now shows tier labels (GOOD / STRONG / EXPERT) instead of raw numbers as you type on ChatGPT, Claude.ai, and Gemini.

IMPROVED: Scoring rebalanced for real-world prompts — clarity and context weighted higher, structure weighted lower. A clear plain-text prompt now scores in the "GOOD" range without requiring markdown formatting.

IMPROVED: Suggestions now show expected point gains (+N pts) so users know which improvements matter most.

Previous features: Real-time score badge, prompt capture, 5-dimension analysis, privacy detection, compression, CLI sync via Native Messaging.

v0.5.0 — Live scoring as you type

New:
• Real-time score badge — see your prompt quality score (0-100) as you type, before sending
• Score updates after 600ms pause, color-coded (green/yellow/red)
• Badge appears near the input box on ChatGPT, Claude.ai, and Gemini

From v0.4.0:
• Repetition detection — orange badge warns when you ask similar questions
• CLI Insights panel — after syncing, see your average score, trend, and tips
• Session grouping — history organized by conversation

All analysis runs locally in your browser. No data leaves your machine.

v0.4.0 — Real-time prompt coaching

New features:
• Real-time repetition detection — orange badge warns when you ask similar questions
• CLI Insights panel — after syncing, see your average score, trend, and coaching tips
• Session grouping — history organized by conversation with per-session stats
• Prompts now show repetition (~) and privacy (⚠) badges in history

Improvements:
• Richer sync protocol — CLI v1.9+ returns insights with every sync
• Insights cached in local storage for persistence across popup opens

All analysis remains 100% local. No data leaves your machine.

v0.3.0: Real-time prompt scoring dashboard with 5-dimension breakdown, privacy detection (warns about API keys/PII before you send), one-click prompt compression & copy to clipboard.

v0.2.1: Updated extension icons — fixed white corner artifacts on store listings, improved visual centering.