Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Multi-repository analysis and research specialist
.claude/skills/bilal140202-researcher/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-09 | ✓→✗ | ▼ Worse | 33% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 22% | 0% |
| case-10 | ✓→✓ | ＝ Same ✓ | 1% | 0% |
You are Researcher, the multi-repository analysis and research specialist.
markdown## Research: Authentication Patterns in Top 100 GitHub Repos ### Methodology - Analyzed top 100 starred Node.js projects - Searched for authentication implementations - Extracted patterns and libraries used ### Findings **Most Common Libraries** (n=100): 1. Passport.js - 42% 2. JWT (jsonwebtoken) - 38% 3. Auth0 SDK - 12% 4. Custom implementation - 8% **Authentication Strategies**: - Local (username/password): 89% - OAuth 2.0 (Google/GitHub): 67% - JWT tokens: 78% - Session-based: 34% **Security Patterns**: - Bcrypt for password hashing: 95% - HTTPS enforcement: 88% - Rate limiting: 72% - CSRF protection: 65% ### Recommendations Based on analysis, recommend: 1. Use Passport.js for flexibility 2. Implement JWT for stateless auth 3. Add OAuth for social login 4. Use bcrypt (cost factor: 12)
Task: Find how top projects handle file uploads
Search Strategy:
1. GitHub code search: "multer upload"
2. Filter by stars > 1000
3. Analyze implementation patterns
Results:
- Express + Multer: Most common (78%)
- File size limits: Usually 10MB
- Storage: S3 (45%), local disk (35%), GCS (20%)
- Validation: File type checking always present
Example from project X:
[Code snippet with best practices]markdown## Feature Comparison: Real-time Frameworks | Feature | Socket.IO | WS | uWebSockets | |---------|-----------|-----|-------------| | WebSocket | ✅ | ✅ | ✅ | | Fallbacks | ✅ | ❌ | ❌ | | Rooms | ✅ | ❌ | ❌ | | Performance | Medium | High | Highest | | Use Case | Full-featured | Simple | High-perf | Recommendation: Socket.IO for most projects - Built-in reconnection - Room support - Proven reliability
"Research is seeing what everybody else has seen and thinking what nobody else has thought." - Albert Szent-Györgyi
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 36,109 | 20,140 | -44% | 1 | 1 | 0% | 6,186 | 3,809 | -38% | 0 | 0 | — |
case-02 | fail→fail | 32,695 | 16,106 | -51% | 1 | 1 | 0% | 3,639 | 3,408 | -6% | 0 | 0 | — |
case-03 | fail→pass | 22,934 | 18,888 | -18% | 1 | 1 | 0% | 3,727 | 3,697 | -1% | 0 | 0 | — |
case-04 | pass→pass | 24,206 | 24,281 | +0% | 1 | 1 | 0% | 3,984 | 4,870 | +22% | 0 | 0 | — |
case-05 | fail→fail | 21,753 | 34,301 | +58% | 1 | 1 | 0% | 3,346 | 4,634 | +38% | 0 | 0 | — |
case-06 | fail→fail | 21,178 | 19,893 | -6% | 1 | 1 | 0% | 3,522 | 4,189 | +19% | 0 | 0 | — |
case-07 | fail→fail | 22,237 | 23,087 | +4% | 1 | 1 | 0% | 3,505 | 4,428 | +26% | 0 | 0 | — |
case-08 | fail→fail | 22,633 | 20,753 | -8% | 1 | 1 | 0% | 3,880 | 4,179 | +8% | 0 | 0 | — |
case-09 | pass→fail | 23,089 | 25,318 | +10% | 1 | 1 | 0% | 3,510 | 4,680 | +33% | 0 | 0 | — |
case-10 | pass→pass | 24,419 | 20,172 | -17% | 1 | 1 | 0% | 3,837 | 3,859 | +1% | 0 | 0 | — |
case-11 | fail→fail | 21,308 | 21,477 | +1% | 1 | 1 | 0% | 3,365 | 3,975 | +18% | 0 | 0 | — |
case-12 | fail→fail | 18,556 | 18,393 | -1% | 1 | 1 | 0% | 3,069 | 3,704 | +21% | 0 | 0 | — |
case-13 | fail→fail | 23,280 | 23,944 | +3% | 1 | 1 | 0% | 3,679 | 4,614 | +25% | 0 | 0 | — |
case-14 | fail→pass | 28,877 | 19,213 | -33% | 1 | 1 | 0% | 4,619 | 3,733 | -19% | 0 | 0 | — |
case-15 | fail→fail | 23,988 | 22,293 | -7% | 1 | 1 | 0% | 3,616 | 4,218 | +17% | 0 | 0 | — |
case-16 | fail→fail | 21,135 | 19,672 | -7% | 1 | 1 | 0% | 3,555 | 3,829 | +8% | 0 | 0 | — |
case-17 | fail→fail | 16,548 | 17,667 | +7% | 1 | 1 | 0% | 2,665 | 3,841 | +44% | 0 | 0 | — |
case-18 | pass→pass | 19,083 | 21,155 | +11% | 1 | 1 | 0% | 3,098 | 3,666 | +18% | 0 | 0 | — |
case-19 | fail→fail | 20,804 | 28,806 | +38% | 1 | 1 | 0% | 3,583 | 5,363 | +50% | 0 | 0 | — |
case-20 | pass→pass | 10,043 | 7,916 | -21% | 1 | 1 | 0% | 1,989 | 2,202 | +11% | 0 | 0 | — |
case-21 | pass→pass | 5,261 | 4,232 | -20% | 1 | 1 | 0% | 834 | 1,429 | +71% | 0 | 0 | — |
case-22 | pass→pass | 11,300 | 12,661 | +12% | 1 | 1 | 0% | 1,920 | 2,880 | +50% | 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. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.