Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Automatically extract reusable patterns from Claude Code sessions and save them as learned skills for future use.
.claude/skills/loulanyue-continuous-learning/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 57% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 21% | 0% |
自動評估 Claude Code 工作階段結束時的內容，提取可重用模式並儲存為學習技能。
此技能作為 Stop hook 在每個工作階段結束時執行：
~/.claude/skills/learned/編輯 config.json 以自訂：
json{ "min_session_length": 10, "extraction_threshold": "medium", "auto_approve": false, "learned_skills_path": "~/.claude/skills/learned/", "patterns_to_detect": [ "error_resolution", "user_corrections", "workarounds", "debugging_techniques", "project_specific" ], "ignore_patterns": [ "simple_typos", "one_time_fixes", "external_api_issues" ] }
| 模式 | 描述 | |------|------| | error_resolution | 特定錯誤如何被解決 | | user_corrections | 來自使用者修正的模式 | | workarounds | 框架/函式庫怪異問題的解決方案 | | debugging_techniques | 有效的除錯方法 | | project_specific | 專案特定慣例 |
新增到你的 ~/.claude/settings.json：
json{ "hooks": { "Stop": [{ "matcher": "*", "hooks": [{ "type": "command", "command": "~/.claude/skills/continuous-learning/evaluate-session.sh" }] }] } }
/learn 指令 - 工作階段中手動提取模式Homunculus v2 採用更複雜的方法：
| 功能 | 我們的方法 | Homunculus v2 | |------|----------|---------------| | 觀察 | Stop hook（工作階段結束） | PreToolUse/PostToolUse hooks（100% 可靠） | | 分析 | 主要上下文 | 背景 agent（Haiku） | | 粒度 | 完整技能 | 原子「本能」 | | 信心 | 無 | 0.3-0.9 加權 | | 演化 | 直接到技能 | 本能 → 聚類 → 技能/指令/agent | | 分享 | 無 | 匯出/匯入本能 |
來自 homunculus 的關鍵見解： > "v1 依賴技能進行觀察。技能是機率性的——它們觸發約 50-80% 的時間。v2 使用 hooks 進行觀察（100% 可靠），並以本能作為學習行為的原子單位。"
參見：docs/continuous-learning-v2-spec.md 完整規格。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-21 | fail→fail | 11,181 | 8,917 | -20% | 1 | 1 | 0% | 2,203 | 2,727 | +24% | 0 | 0 | — |
case-07 | pass→pass | 14,761 | 11,658 | -21% | 1 | 1 | 0% | 2,190 | 2,836 | +29% | 0 | 0 | — |
case-08 | pass→pass | 17,030 | 17,592 | +3% | 1 | 1 | 0% | 2,575 | 3,648 | +42% | 0 | 0 | — |
case-01 | fail→pass | 16,303 | 7,851 | -52% | 1 | 1 | 0% | 2,923 | 2,404 | -18% | 0 | 0 | — |
case-02 | fail→pass | 31,983 | 7,249 | -77% | 1 | 1 | 0% | 2,528 | 2,356 | -7% | 0 | 0 | — |
case-03 | fail→pass | 18,010 | 8,159 | -55% | 1 | 1 | 0% | 3,224 | 2,412 | -25% | 0 | 0 | — |
case-04 | pass→pass | 12,422 | 14,279 | +15% | 1 | 1 | 0% | 1,830 | 3,094 | +69% | 0 | 0 | — |
case-05 | pass→pass | 15,871 | 12,358 | -22% | 1 | 1 | 0% | 2,349 | 2,817 | +20% | 0 | 0 | — |
case-06 | pass→pass | 13,013 | 10,230 | -21% | 1 | 1 | 0% | 1,901 | 2,514 | +32% | 0 | 0 | — |
case-09 | fail→pass | 12,052 | 10,925 | -9% | 1 | 1 | 0% | 1,681 | 2,646 | +57% | 0 | 0 | — |
case-10 | fail→pass | 12,803 | 2,059 | -84% | 1 | 1 | 0% | 1,136 | 1,370 | +21% | 0 | 0 | — |
case-11 | pass→pass | 16,236 | 12,927 | -20% | 1 | 1 | 0% | 2,253 | 2,995 | +33% | 0 | 0 | — |
case-12 | pass→pass | 14,871 | 14,233 | -4% | 1 | 1 | 0% | 2,377 | 3,311 | +39% | 0 | 0 | — |
case-13 | pass→pass | 14,869 | 16,619 | +12% | 1 | 1 | 0% | 2,191 | 3,454 | +58% | 0 | 0 | — |
case-14 | pass→pass | 14,683 | 14,449 | -2% | 1 | 1 | 0% | 2,138 | 3,166 | +48% | 0 | 0 | — |
case-15 | pass→pass | 34,657 | 16,719 | -52% | 1 | 1 | 0% | 2,818 | 3,470 | +23% | 0 | 0 | — |
case-16 | fail→pass | 15,887 | 12,268 | -23% | 1 | 1 | 0% | 2,250 | 2,818 | +25% | 0 | 0 | — |
case-17 | pass→pass | 9,783 | 9,211 | -6% | 1 | 1 | 0% | 1,748 | 2,500 | +43% | 0 | 0 | — |
case-18 | pass→pass | 11,022 | 6,724 | -39% | 1 | 1 | 0% | 1,690 | 2,064 | +22% | 0 | 0 | — |
case-19 | pass→pass | 12,100 | 11,380 | -6% | 1 | 1 | 0% | 1,766 | 2,702 | +53% | 0 | 0 | — |
case-20 | fail→fail | 30,660 | 2,731 | -91% | 1 | 1 | 0% | 1,101 | 1,328 | +21% | 0 | 0 | — |
case-22 | fail→pass | 8,193 | 6,324 | -23% | 1 | 1 | 0% | 1,323 | 1,916 | +45% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +32 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.