bybayramannakov· 33 skills
Self-learning system that captures corrections during sessions and reminds users to run /reflect to update CLAUDE.md. Use when discussing learnings, corrections, or when the user mentions remembering something for future sessions.
$npx -y skills add bayramannakov/claude-reflect --skill claude-reflectInstalls into the current project.
Run `npx skills use "https://github.com/bayramannakov/claude-reflect" --skill "bayramannakov/claude-reflect"` and follow the generated skill instructions now. Read its complete output, redirecting it to a temporary file first if necessary. Resolve relative paths from the supporting-files directory it provides.
Use the skills in "https://github.com/bayramannakov/claude-reflect" that are relevant to the current task. Run `npx skills add "https://github.com/bayramannakov/claude-reflect"` and select the relevant skills, then follow their instructions.
| 1 | # claude-reflect |
| 2 | |
| 3 | [](https://github.com/BayramAnnakov/claude-reflect/stargazers) |
| 4 | [](https://github.com/BayramAnnakov/claude-reflect/releases) |
| 5 | [](https://opensource.org/licenses/MIT) |
| 6 | [](https://github.com/BayramAnnakov/claude-reflect/actions) |
| 7 | [](https://github.com/BayramAnnakov/claude-reflect#platform-support) |
| 8 | |
| 9 | A self-learning system for Claude Code that captures corrections and discovers workflow patterns — turning them into permanent memory and reusable skills. |
| 10 | |
| 11 | ## What it does |
| 12 | |
| 13 | ### 1. Learn from Corrections |
| 14 | |
| 15 | When you correct Claude ("no, use gpt-5.1 not gpt-5"), it remembers forever. |
| 16 | |
| 17 | ``` |
| 18 | ┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐ |
| 19 | │ You correct │ ──► │ Hook captures │ ──► │ /reflect adds │ |
| 20 | │ Claude Code │ │ to queue │ │ to CLAUDE.md │ |
| 21 | └─────────────────┘ └─────────────────┘ └─────────────────┘ |
| 22 | (automatic) (automatic) (manual review) |
| 23 | ``` |
| 24 | |
| 25 | ### 2. Discover Workflow Patterns (NEW in v2) |
| 26 | |
| 27 | Analyzes your session history to find repeating tasks that could become reusable commands. |
| 28 | |
| 29 | ``` |
| 30 | ┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐ |
| 31 | │ Your past │ ──► │ /reflect-skills │ ──► │ Generates │ |
| 32 | │ sessions │ │ finds patterns │ │ /commands │ |
| 33 | └─────────────────┘ └─────────────────┘ └─────────────────┘ |
| 34 | (68 sessions) (AI-powered) (you approve) |
| 35 | ``` |
| 36 | |
| 37 | Example: You've asked "review my productivity" 12 times → suggests creating `/daily-review` |
| 38 | |
| 39 | ## Key Features |
| 40 | |
| 41 | | Feature | What it does | |
| 42 | |---------|--------------| |
| 43 | | **Permanent Memory** | Corrections sync to CLAUDE.md — Claude remembers across sessions | |
| 44 | | **Skill Discovery** | Finds repeating patterns in your history → generates commands | |
| 45 | | **Multi-language** | AI understands corrections in any language | |
| 46 | | **Skill Improvement** | Corrections during `/deploy` improve the deploy skill itself | |
| 47 | |
| 48 | ## Installation |
| 49 | |
| 50 | ```bash |
| 51 | # Add the marketplace |
| 52 | claude plugin marketplace add bayramannakov/claude-reflect |
| 53 | |
| 54 | # Install the plugin |
| 55 | claude plugin install claude-reflect@claude-reflect-marketplace |
| 56 | |
| 57 | # IMPORTANT: Restart Claude Code to activate the plugin |
| 58 | ``` |
| 59 | |
| 60 | After installation, **restart Claude Code** (exit and reopen). Then hooks auto-configure and commands are ready. |
| 61 | |
| 62 | > **First run?** When you run `/reflect` for the first time, you'll be prompted to scan your past sessions for learnings. |
| 63 | |
| 64 | ### Prerequisites |
| 65 | |
| 66 | - [Claude Code](https://claude.ai/code) CLI installed |
| 67 | - Python 3.6+ (included on most systems) |
| 68 | |
| 69 | ### Platform Support |
| 70 | |
| 71 | - **macOS**: Fully supported |
| 72 | - **Linux**: Fully supported |
| 73 | - **Windows**: Fully supported (native Python, no WSL required) |
| 74 | |
| 75 | ## Commands |
| 76 | |
| 77 | | Command | Description | |
| 78 | |---------|-------------| |
| 79 | | `/reflect` | Process queued learnings with human review | |
| 80 | | `/reflect --scan-history` | Scan ALL past sessions for missed learnings | |
| 81 | | `/reflect --dry-run` | Preview changes without applying | |
| 82 | | `/reflect --targets` | Show detected config files (CLAUDE.md, AGENTS.md) | |
| 83 | | `/reflect --review` | Show queue with confidence scores and decay status | |
| 84 | | `/reflect --dedupe` | Find and consolidate similar entries in CLAUDE.md | |
| 85 | | `/reflect --include-tool-errors` | Include tool execution errors in scan | |
| 86 | | `/reflect-skills` | Discover skill candidates from repeating patterns | |
| 87 | | `/reflect-skills --days N` | Analyze last N days (default: 14) | |
| 88 | | `/reflect-skills --project <path>` | Analyze specific project | |
| 89 | | `/reflect-skills --all-projects` | Scan all projects for cross-project patterns | |
| 90 | | `/reflect-skills --dry-run` | Preview patterns without generating skill files | |
| 91 | | `/skip-reflect` | Discard all queued learnings | |
| 92 | | `/view-queue` | View pending learnings without processing | |
| 93 | |
| 94 | ## How It Works |
| 95 | |
| 96 |  |
| 97 | |
| 98 | ### Two-Stage Process |
| 99 | |
| 100 | **Stage 1: Capture (Automatic)** |
| 101 | |
| 102 | Hooks run automatically to detect and queue corrections: |
| 103 | |
| 104 | | Hook | Trigger | Purpose | |
| 105 | |------|---------|---------| |
| 106 | | `session_start_reminder.py` | Session start | Shows pending learnings reminder | |
| 107 | | `capture_learning.py` | Every prompt | Detects correction patterns and queues them | |
| 108 | | `check_learnings.py` | Before compaction | Backs up queue and informs user | |
| 109 | | `post_commit_reminder.py` | After git commit | Reminds to run /reflect after completing work | |
| 110 | |
| 111 | **Stage 2: Process (Manual)** |
| 112 | |
| 113 | Run `/reflect` to review and apply queued learnings to CLAUDE.md. |
| 114 | |
| 115 | ### Detection Methods |
| 116 | |
| 117 | Claude-reflect uses a **hybrid detection approach**: |
| 118 | |
| 119 | **1. Regex patterns (real-time capture)* |