[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-81831":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":14,"stars7d":15,"stars30d":15,"stars90d":14,"forks30d":14,"starsTrendScore":14,"compositeScore":16,"rankGlobal":9,"rankLanguage":9,"license":17,"archived":18,"fork":18,"defaultBranch":19,"hasWiki":20,"hasPages":18,"topics":21,"createdAt":9,"pushedAt":9,"updatedAt":22,"readmeContent":23,"aiSummary":24,"trendingCount":14,"starSnapshotCount":14,"syncStatus":12,"lastSyncTime":25,"discoverSource":26},81831,"Medusa","jtshow\u002FMedusa","jtshow","Medusa Skill Framework for AI Agents.",null,"Rust",32,2,1,0,3,1.43,"Other",false,"main",true,[],"2026-06-12 02:04:20","# Medusa Skill Framework (MSF) v0.12.1\n[![medusa.jpg](https:\u002F\u002Fi.postimg.cc\u002Fk5v9H1qt\u002Fmedusa.jpg)](https:\u002F\u002Fpostimg.cc\u002Fp9rghB92)\n[![W6EJw.jpg](https:\u002F\u002Fi.postimg.cc\u002Fk4bCxgCN\u002FW6EJw.jpg)](https:\u002F\u002Fpostimg.cc\u002FR642m9Nq)\n\nUltra-fast skill scanner with **audit-based ranking**, automatic promotion, and 9-tier leveling system.\n\n## ⚡ One-Line Install\n\n### Windows (PowerShell)\n```powershell\nirm https:\u002F\u002Fraw.githubusercontent.com\u002Fyour-repo\u002Fmedusa\u002Fmain\u002Finstall.ps1 | iex\n```\n\n### Windows (Command Prompt)\n```batch\ncurl -SL https:\u002F\u002Fraw.githubusercontent.com\u002Fyour-repo\u002Fmedusa\u002Fmain\u002Finstall.bat -o install.bat && install.bat\n```\n\n### macOS \u002F Linux\n```bash\ncurl -sSL https:\u002F\u002Fraw.githubusercontent.com\u002Fyour-repo\u002Fmedusa\u002Fmain\u002Finstall.sh | bash\n```\n\n### Build from Source (Any Platform)\n\n**Step 1: Install Rust**\n- Windows: `irm https:\u002F\u002Fwin.rustup.rs\u002Fx86_64 | iex` (or download from https:\u002F\u002Frustup.rs)\n- macOS\u002FLinux: `curl --proto '=https' --tlsv1.2 -sSf https:\u002F\u002Fsh.rustup.rs | sh`\n\n**Step 2: Clone & Build**\n```bash\n# All platforms (Windows CMD\u002FPowerShell, macOS, Linux):\ngit clone https:\u002F\u002Fgithub.com\u002Fyour-repo\u002Fmedusa.git\ncd medusa\ncargo build --release\n```\n\n**Binary Location:**\n| Platform | Binary Path |\n|----------|--------------|\n| **Windows** | `target\\release\\medusa.exe` |\n| **macOS\u002FLinux** | `target\u002Frelease\u002Fmedusa` |\n\n**Step 3: Run**\n```bash\n# Windows (CMD)\n.\\target\\release\\medusa.exe --help\n\n# Windows (PowerShell)\n.\\target\\release\\medusa.exe --help\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa --help\n```\n\n| Feature | Description | Performance |\n|---------|-------------|-------------|\n| **Audit-Based Ranking** | Measures complexity, value, technical depth | 9-tier system |\n| **Automatic Promotion** | Skills rank up as they improve | No manual needed |\n| **Parallel Scanning** | Rayon-powered concurrent processing | 46% faster (A\u002FB tested) |\n| **Fusion Detection** | Finds similar skills (name + content) | FxHash-powered |\n| **HTML Visualization** | Dark-themed reports with progress bars | Interactive |\n| **A\u002FB Test Framework** | Validate performance claims scientifically | Statistical rigor |\n| **Dreaming Process** | Cross-session pattern detection, recurring gaps, trends | Auto-records every scan |\n| **Memory Consolidation** | Merges duplicates, prunes low-severity, caps at 200 | Runs after every dream |\n| **Outcomes Framework** | Weighted rubric-based skill assessment | 4 default criteria |\n| **Learning Paths** | Built-in fix suggestions mapped to gap patterns | Per-skill recommendations |\n| **Multi-Agent Orchestration** | 4 specialized sub-audits (doc quality, code quality, dependencies, learning value) | Synthesized weighted scoring |\n| **Dream Diary** | Narrative timeline of skill evolution with gap history | Console + Markdown export |\n| **Configurable Dreaming** | Tune frequency, retention, auto-apply, max insights | Via medusa.toml |\n| **Procedural Memory** | Auto-detects step-by-step workflows from skill content | 38 workflows from 36 skills |\n| **Cross-Agent Memory** | Export\u002Fimport dream + procedural + outcome bundles | Merge with source tracking |\n\n## Usage\n\n### Scan Skills (JSON Output)\n\n```bash\n# Windows (CMD)\n.\\target\\release\\medusa.exe scan C:\\path\\to\\skills\n\n# Windows (PowerShell)\n.\\target\\release\\medusa.exe scan C:\\path\\to\\skills\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa scan \u002Fpath\u002Fto\u002Fskills\n```\n\n### Audit a Skill (See WHY it's at its tier)\n\n```bash\n# Windows\n.\\target\\release\\medusa.exe audit C:\\path\\to\\skills\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa audit \u002Fpath\u002Fto\u002Fskills\n```\n\n**Example Output:**\n```\n=== Medusa Skill Audit Report ===\n\nSkill: ai-ml (ai-ml), level: Godlike\n  Experience: 100.0\u002F100\n  Confidence: 75%\n  Metrics:\n    - Content Length: 5966 chars\n    - Code Blocks: 15\n    - Step Instructions: 0\n    - Technical Terms: 26\n    - Complexity Score: 80.0\u002F100\n    - Value Score: 90.0\u002F100\n```\n\n### Generate HTML Report\n\n```bash\n# Windows (CMD)\n.\\target\\release\\medusa.exe html C:\\path\\to\\skills C:\\path\\to\\report.html\n\n# Windows (PowerShell)\n.\\target\\release\\medusa.exe html C:\\path\\to\\skills C:\\path\\to\\report.html\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa html \u002Fpath\u002Fto\u002Fskills \u002Fpath\u002Fto\u002Freport.html\n```\n\nOpens a beautiful dark-themed visualization with:\n- Skill bars showing experience levels\n- Color-coded tiers (Godlike = Purple gradient, Unique = Orange gradient, etc.)\n- Detailed metrics for each skill\n- Fusion detection (similar skills)\n\n### Run A\u002FB Test (Validate Performance)\n\n```bash\n# Windows (CMD)\n.\\target\\release\\medusa.exe ab-test C:\\path\\to\\skills --iterations 20\n\n# Windows (PowerShell)\n.\\target\\release\\medusa.exe ab-test C:\\path\\to\\skills --iterations 20\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa ab-test \u002Fpath\u002Fto\u002Fskills --iterations 20\n```\n\n**Example Output:**\n```\nRunning A\u002FB Test: Parallel vs Sequential Scan\nPath: \u002Fpath\u002Fto\u002Fskills\nIterations: 20\n\nHypothesis: Parallel scanning is faster than sequential\nPrimary metric: scan_time_ms\n\nIteration 1: Parallel=178ms, Sequential=362ms\nIteration 2: Parallel=187ms, Sequential=325ms\n...\nIteration 20: Parallel=204ms, Sequential=391ms\n\n=== A\u002FB Test Results ===\nParallel avg: 190.00ms\nSequential avg: 352.00ms\n✅ Parallel is 46.0% faster\n```\n\n## 9-Tier Leveling System\n\n| Tier | Range | Color | Background |\n|------|--------|--------|-------------|\n| **Godlike** | 95+ | 🔴 Red-Orange-Green | Gradient |\n| **Unique** | 90+ | 🔴 Red | Solid |\n| **Legendary** | 85+ | 💜 Pink-Purple | Solid |\n| **Mythic** | 80+ | 🟣 Purple | Solid |\n| **Epic** | 75+ | 🟡 Yellow | Solid |\n| **Ultra Rare** | 65+ | 🟢 Teal | Solid |\n| **Rare** | 55+ | 🔵 Blue | Solid |\n| **Uncommon** | 45+ | 🟢 Green | Solid |\n| **Common** | 25+ | ⚪ Light Gray | Solid |\n| **Poor** | \u003C25 | ⚫ Dark Gray | Solid |\n\n## Commands\n\n```bash\nmedusa --help\n```\n\nOutput:\n```\nMedusa Skill Framework (MSF) v0.12 - Audit-Based Ranking with Context\nUsage: medusa \u003Ccommand> [options]\n\nCommands:\n  scan \u003Cpath>              Scan skills with FULL audit (60\u002F30\u002F10 scoring)\n    --sequential           Use sequential scanning (no Rayon)\n    --no-cache             Disable incremental scan cache\n\n  audit \u003Cpath>             Show detailed skill audit with cross-session context\n    --no-cache             Disable cache\n\n  html \u003Cpath> \u003Coutput>     Generate HTML visualization\n    --sequential           Use sequential scanning\n    --no-cache             Disable cache\n\n  export-csv \u003Cpath> \u003Cf>    Export skills to CSV format\n  export-md \u003Cpath> \u003Cf>     Export skills to Markdown\n  export-svg \u003Cpath> \u003Cf>    Export skills to SVG visualization\n\n  ab-test \u003Cpath>           Run A\u002FB test (parallel vs sequential)\n    --iterations N         Number of test iterations (default: 10)\n\n  dream \u003Cpath>             Run dreaming process (cross-session pattern detection)\n  dream-status \u003Cpath>      Show dream knowledge base and patterns\n  dream-reset \u003Cpath>       Reset dream state and history\n  dream-consolidate \u003Cpath> Manually consolidate dream knowledge base\n  dream-diary \u003Cpath>       Show dream diary (narrative skill evolution timeline)\n    --output \u003Cfile.md>     Export diary as Markdown\n  dream-params \u003Cpath>      Show dreaming configuration parameters\n\n  orchestrate \u003Cpath>       Run multi-agent orchestrated audit (4 specialized sub-audits)\n    --sequential           Use sequential scanning\n    --no-cache             Disable cache\n\n  outcome-add \u003Cpath> \u003Cid>  Add default outcome rubric for a skill\n  outcome-list \u003Cpath>      List outcome rubrics\n  outcome-remove \u003Cpath> \u003Cid> Remove an outcome rubric\n  learning-path \u003Cpath> \u003Cid> Show learning path and suggestions for a skill\n\n  procedural-list \u003Cpath>   List all learned procedural workflows\n  procedural-show \u003Cp> \u003Cid> Show workflows associated with a skill\n\n  memory-export \u003Cp> \u003Cf>    Export all memory (dream, procedural, outcomes) to JSON bundle\n  memory-import \u003Cp> \u003Cf>    Import and merge a memory bundle from another Medusa instance\n    --source \u003Cname>        Tag imported data with a source identifier\n\n  update                  Update Medusa from GitHub (git pull + rebuild)\n\nOptions:\n  --help, -h              Show this help message\n  --version, -v           Show version\n```\n\n## How It Works\n\n```\nSKILL.md files\n    ↓\n[WalkDir] Scan filesystem (max depth 4)\n    ↓\n[Rayon] Parallel processing (46% faster, optional)\n    ↓\n[Regex] Extract YAML frontmatter\n    ↓\n[Audit] Measure complexity (length, code, steps, terms)\n    ↓\n[Score] Calculate experience (60% + 30% + 10%)\n    ↓\n[Rank] Assign tier (Godlike → Poor)\n    ↓\n[Fusion] Detect similar skills (FxHash)\n    ↓\n[Session] Record snapshot for dreaming\n    ↓\n[Dream] Cross-session pattern detection + consolidation\n    ↓\n[Diary] Generate skill evolution timeline\n    ↓\n[Agents] Multi-agent orchestrated sub-audits\n    ↓\n[Outcomes] Rubric-based quality assessment\n    ↓\n[Procedural] Extract step-by-step workflows\n    ↓\n[Output] JSON \u002F HTML \u002F CSV \u002F MD \u002F SVG\n```\n\n## Cross-Platform Support ✅\n\n| Platform | Binary | Build Command |\n|----------|--------|---------------|\n| **Windows** | `medusa.exe` | `cargo build --release` |\n| **macOS (Intel)** | `medusa` | `cargo build --release` |\n| **macOS (Apple Silicon)** | `medusa` | `cargo build --release` |\n| **Linux (x86_64)** | `medusa` | `cargo build --release` |\n| **Linux (ARM64)** | `medusa` | `cargo build --release` |\n\n**No WSL required!** Runs natively on all platforms.\n\n```\nmedusa\u002F\n├── src\u002F\n│   ├── main.rs          # CLI entry, scoring, reports (~1350 lines)\n│   ├── dream.rs         # Dreaming, consolidation, diary, learning paths\n│   ├── outcomes.rs      # Rubric CRUD, skill assessment\n│   ├── agents.rs        # Multi-agent orchestrated audit (4 agents)\n│   └── procedural.rs    # Procedural workflow detection & memory\n├── target\u002F\n│   └── release\u002F\n│       └── medusa       # Compiled binary\n├── Cargo.toml          # Dependencies (minimal: 9 deps)\n├── medusa.toml         # Configurable scoring + dreaming params\n├── README.md           # This file\n├── TECHNICAL.md        # Architecture deep-dive\n└── .medusa_state.json  # Promotion state (auto-created)\n```\n\n## Dependencies (Minimal)\n\n```\nserde       — Struct serialization\nserde_json  — JSON output\ntoml        — Config parsing\nfxhash      — Fusion hash computation\nwalkdir     — Directory traversal\nrayon       — Parallel processing\nregex       — Pattern extraction\nlazy_static — Regex compilation\nchrono      — Timestamps for dream sessions\n```\n\n**Compile time**: ~5 seconds (release, stripped)\n**Binary size**: ~2MB (Windows\u002FLinux\u002FmacOS)\n**No runtime dependencies!** Single binary, just download and run.\n\n## Examples\n\n### Example 1: Scan Your Skills\n\n```bash\n# Windows\n.\\target\\release\\medusa.exe scan C:\\Project\\.opencode\\skills\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa scan ~\u002F.hermes\u002Fskills\n```\n\n### Example 2: Audit a Specific Skill\n\n```bash\n# Windows\n.\\target\\release\\medusa.exe audit C:\\Project\\.opencode\\skills\\ai-ml\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa audit ~\u002F.hermes\u002Fskills\u002Fai-ml\n```\n\nShows detailed breakdown:\n- Why it's ranked \"Godlike\"\n- Content length, code blocks, step count\n- Technical term density\n- Complexity and value scores\n\n### Example 3: Generate Beautiful Report\n\n```bash\n# Windows\n.\\target\\release\\medusa.exe html C:\\Project\\.opencode\\skills C:\\Project\\medusa\\report.html\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa html ~\u002F.hermes\u002Fskills ~\u002Freport.html\n```\n\n### Example 2: Audit a Specific Skill\n```bash\n# Shows WHY it's at its tier\nmedusa audit \u002Fpath\u002Fto\u002Fskills\u002Fai-ml\n```\n\n**Example Output:**\n```\n=== Medusa Skill Audit Report ===\n\nSkill: ai-ml (ai-ml), level: Godlike\n  Experience: 100.0\u002F100\n  Confidence: 75%\n  Metrics:\n    - Content Length: 5966 chars\n    - Code Blocks: 15\n    - Step Instructions: 0\n    - Technical Terms: 26\n    - Complexity Score: 80.0\u002F100\n    - Value Score: 90.0\u002F100\n```\n\n### Example 3: Generate Beautiful Report\n```bash\n# Windows\n.\\target\\release\\medusa.exe html C:\\path\\to\\skills report.html\n\n# Mac\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa html \u002Fpath\u002Fto\u002Fskills report.html\n```\n\nOpens in browser with:\n- Dark theme (hacker style)\n- Color-coded skill cards\n- Experience progress bars\n- Fusion detection section\n\n## Performance\n\n**A\u002FB Test Results** (36 skills, 20 iterations):\n- **Parallel** (Rayon): 190ms average\n- **Sequential**: 352ms average\n- **Speedup**: 46% faster\n\n**Scalability**:\n- 36 skills scanned in ~150ms\n- Linear scaling with Rayon parallelization\n- Memory-efficient (no unnecessary copies)\n\n## Update Medusa\n\n```bash\n# One-line update (pull latest + rebuild)\ncd \u002Fpath\u002Fto\u002Fmedusa && git pull && cargo build --release\n```\n\nOr use the built-in update command:\n```bash\nmedusa update\n```\n\n## Quick Test (Verify Installation)\n\nAfter installation, verify it works:\n\n```bash\n# Windows (CMD)\n.\\target\\release\\medusa.exe --version\n\n# Windows (PowerShell)\n.\\target\\release\\medusa.exe --version\n\n# macOS\u002FLinux\n.\u002Ftarget\u002Frelease\u002Fmedusa --version\n```\n\n**Expected output:**\n```\nMedusa Skill Framework (MSF) v0.12.1\n```\n\n### Test Scan\n```bash\n# Windows: medusa scan C:\\Project\\.opencode\\skills\n\n# macOS\u002FLinux:\n.\u002Ftarget\u002Frelease\u002Fmedusa scan ~\u002F.hermes\u002Fskills\n```\n\nShould output JSON with skills audit.\n\nTraditional skill systems use **static rankings** (you manually set the level).\n\nMedusa uses **audit-based ranking**:\n1. **Measures** actual skill complexity (content, code, steps, terms)\n2. **Calculates** objective experience score (60% complexity + 30% value)\n3. **Assigns** tier automatically (Godlike → Poor)\n4. **Promotes** as you improve (just edit SKILL.md, next scan updates!)\n\n**No manual promotion commands needed!**\n\n## License\n\nMIT License (or your choice)\n\n## Version History\n\n- **v0.12** (Current): Multi-agent orchestration, dream diary, configurable dreaming, procedural memory, cross-agent memory sharing + all Phase 1 features\n- **v0.11**: 9-tier leveling system (Godlike → Poor), rank promotion system\n- **v0.5.0**: Rank promotion system\n- **v0.4.0**: CLI improvements, A\u002FB test framework\n- **v0.3.0**: Fusion detection, HTML visualization\n- **v0.2.0**: Parallel scanning with Rayon\n- **v0.1.0**: Initial release\n\n---\n\n**Built with Rust 🦀 + Rayon ⚡ + Regex 🔍**\n","Medusa是一个专为AI代理设计的技能框架，采用Rust语言编写。它提供了一个基于审计的排名系统，能够自动提升技能等级，并通过9级分级体系对技能进行评估。此外，Medusa还支持并行扫描、融合检测以及HTML可视化报告等功能，显著提升了处理速度和用户体验。特别适用于需要快速分析与优化AI技能集的应用场景中，如自动化测试环境、持续集成\u002F持续部署（CI\u002FCD）流程等。其强大的数据处理能力和灵活配置选项使其成为开发人员和研究者理想的工具选择。","2026-06-11 04:06:52","CREATED_QUERY"]