Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Fetch up-to-date library documentation via Context7 API. Use PROACTIVELY when: (1) Working with ANY external library (React, Next.js, Supabase, etc.) (2) User asks about library APIs, patterns, or best practices (3) Implementing features that rely on third-party packages (4) Debugging library-specific issues (5) Need current documentation beyond training data cutoff (6) AND MOST IMPORTANTLY, when you are installing dependencies, libraries, or frameworks you should ALWAYS check the docs to see wh
.claude/skills/aiskillstore-context7/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 54% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 174% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 190% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 271% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 36% | 0% |
Search GitHub repositories for issues, PRs, discussions, and code examples to research solutions and best practices.
| 路径类型 | 说明 | |---------|------| | 使用方式 | 此技能通过 API 调用使用,无需本地脚本执行 | | 调用场景 | 当用户需要搜索 GitHub 仓库时自动激活 | | 输入参数 | 仓库名称 (owner/repo)、搜索关键词、过滤条件 |
owner/repo(如 ggerganov/whisper.cpp)VoiceLite dependencies and related projects:
| Repository | Purpose | When to Search | |------------|---------|----------------| | ggerganov/whisper.cpp | Core transcription engine | Performance optimization, model loading, quantization issues | | naudio/NAudio | Audio recording library | WaveInEvent issues, audio format problems, disposal patterns | | dotnet/wpf | WPF framework | UI threading, XAML binding, Dispatcher issues | | jrsoftware/issrc | Inno Setup installer | Installer configuration, file inclusion, signing | | dotnet/runtime | .NET runtime | Performance issues, GC problems, async/await patterns |
Repository: ggerganov/whisper.cpp
Query: "performance optimization" label:performance
Sort: Most commented
Filter: Created after 2024-01-01
# Look for:
- Quantization discussions (Q8_0, Q4_0)
- Flash attention implementations
- Beam size optimization
- Model loading speed improvementsRepository: naudio/NAudio
Query: "WaveInEvent" label:bug is:closed
Sort: Recently updated
# Look for:
- Disposal patterns (memory leaks)
- Buffer size configurations
- Sample rate issues (16kHz mono)
- Event subscription patternsRepository: dotnet/wpf
Query: "Dispatcher.Invoke" in:code language:csharp
Filter: Stars >100
# Look for:
- Thread-safe UI updates
- Background worker patterns
- Async dispatcher usageStep 1: Search issue titles
→ "transcription slow"
Step 2: Add labels
→ "transcription slow" label:performance
Step 3: Check discussions
→ Switch to Discussions tab for detailed solutions
Step 4: Look at closed issues
→ is:closed (solutions often in closed issues)For bugs:
For features:
# Search for actual code implementation
in:code language:csharp "WaveInEvent"
# Search for configuration examples
in:file filename:.csproj "NAudio"
# Search for specific patterns
in:code "async Task TranscribeAsync"Query: "Q8_0 quantization" OR "performance improvement"
Repo: ggerganov/whisper.cpp
Labels: performance, optimization
Date: After 2024-01-01
Expected: Quantization benchmarks, speed comparisons, optimization tipsQuery: "memory leak" OR "dispose" "WaveInEvent"
Repo: naudio/NAudio
State: Closed (to find fixes)
Sort: Most commented
Expected: Disposal patterns, IDisposable best practicesQuery: "files not included" OR "missing from installer"
Repo: jrsoftware/issrc
Labels: bug, question
Expected: Common .iss mistakes, file path issues, git ignore problemsQuery: "Process.Kill" OR "zombie process"
Repo: dotnet/runtime
Language: C#
Expected: Proper disposal patterns, timeout handling# Combine multiple terms
"whisper performance" AND "quantization"
# Exclude terms
"audio recording" NOT "streaming"
# Search specific user
author:ggerganov "optimization"
# Search by date range
created:>=2024-01-01
# Search by reactions
reactions:>10
# Search by comments
comments:>5
# Search in specific locations
in:title "memory leak"
in:body "WaveInEvent"
in:comments "fixed in"Scenario: VoiceLite transcription is slow with tiny model
Step 1: Search whisper.cpp issues
→ Query: "tiny model slow" label:performance
→ Find: Issue #1234 - "Tiny model slower than expected"
Step 2: Read discussion
→ Solution: Enable flash attention, adjust beam size
→ PR #5678 has implementation
Step 3: Check PR for code changes
→ Command line flag: --flash-attn
→ Configuration: beam_size=1
Step 4: Check if applied to VoiceLite
→ Review PersistentWhisperService.cs whisper command
→ Verify flags are present
Step 5: Test & validate
→ Apply if missing, test performance improvementis:closed for solved problemsWhen researching VoiceLite issues, search these patterns:
Audio Issues: NAudio + "16kHz" OR "mono" OR "WAV format" Transcription Issues: whisper.cpp + "model loading" OR "timeout" OR "process" Performance Issues: whisper.cpp + "Q8_0" OR "optimization" OR "speed" Installer Issues: Inno Setup + "missing files" OR "not included" Memory Issues: .NET + "memory leak" OR "dispose" OR "GC"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 27,675 | 19,765 | -29% | 1 | 1 | 0% | 2,635 | 4,069 | +54% | 0 | 0 | — |
case-02 | fail→pass | 10,307 | 10,388 | +1% | 1 | 1 | 0% | 942 | 2,578 | +174% | 0 | 0 | — |
case-03 | fail→pass | 24,974 | 11,160 | -55% | 1 | 1 | 0% | 919 | 2,669 | +190% | 0 | 0 | — |
case-04 | fail→fail | 13,150 | 6,340 | -52% | 1 | 1 | 0% | 1,243 | 2,707 | +118% | 0 | 0 | — |
case-05 | pass→pass | 8,156 | 12,827 | +57% | 1 | 1 | 0% | 1,285 | 2,820 | +119% | 0 | 0 | — |
case-06 | pass→pass | 10,789 | 33,758 | +213% | 1 | 1 | 0% | 1,673 | 3,010 | +80% | 0 | 0 | — |
case-07 | pass→pass | 8,239 | 3,693 | -55% | 1 | 1 | 0% | 416 | 2,267 | +445% | 0 | 0 | — |
case-08 | fail→fail | 8,536 | 8,480 | -1% | 1 | 1 | 0% | 532 | 2,260 | +325% | 0 | 0 | — |
case-09 | pass→pass | 4,137 | 2,490 | -40% | 1 | 1 | 0% | 563 | 2,039 | +262% | 0 | 0 | — |
case-10 | fail→fail | 8,449 | 9,690 | +15% | 1 | 1 | 0% | 1,505 | 2,508 | +67% | 0 | 0 | — |
case-11 | fail→pass | 8,822 | 9,143 | +4% | 1 | 1 | 0% | 618 | 2,292 | +271% | 0 | 0 | — |
case-12 | pass→pass | 11,010 | 8,797 | -20% | 1 | 1 | 0% | 990 | 2,313 | +134% | 0 | 0 | — |
case-13 | pass→pass | 12,580 | 8,188 | -35% | 1 | 1 | 0% | 1,810 | 2,976 | +64% | 0 | 0 | — |
case-14 | pass→pass | 9,467 | 8,068 | -15% | 1 | 1 | 0% | 636 | 2,173 | +242% | 0 | 0 | — |
case-15 | pass→pass | 9,198 | 7,834 | -15% | 1 | 1 | 0% | 508 | 2,100 | +313% | 0 | 0 | — |
case-16 | fail→pass | 16,209 | 12,453 | -23% | 1 | 1 | 0% | 2,680 | 3,639 | +36% | 0 | 0 | — |
case-17 | pass→pass | 12,294 | 12,142 | -1% | 1 | 1 | 0% | 2,079 | 3,166 | +52% | 0 | 0 | — |
case-18 | fail→pass | 18,849 | 8,526 | -55% | 1 | 1 | 0% | 2,078 | 2,624 | +26% | 0 | 0 | — |
case-19 | fail→pass | 14,110 | 11,645 | -17% | 1 | 1 | 0% | 1,390 | 2,780 | +100% | 0 | 0 | — |
case-20 | fail→pass | 13,216 | 3,309 | -75% | 1 | 1 | 0% | 1,275 | 2,209 | +73% | 0 | 0 | — |
case-21 | pass→pass | 18,548 | 13,896 | -25% | 1 | 1 | 0% | 2,906 | 4,889 | +68% | 0 | 0 | — |
case-22 | pass→pass | 18,570 | 18,407 | -1% | 1 | 1 | 0% | 2,575 | 5,135 | +99% | 0 | 0 | — |
case-23 | pass→pass | 9,078 | 8,528 | -6% | 1 | 1 | 0% | 685 | 2,330 | +240% | 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. 23 cases were attempted, and 22 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 +35 percentage points is the difference between those two pass rates over the 22 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.