Insights

Key findings from analyzing 8,517 popular open-source repositories.

79%
F-grade rate
40
average score
30
security avg
61
Rust avg

Most repos are in poor health

79% of graded repositories (6117 of 7715) received an F grade, making it the most common grade. Only 1% (39) earned an A. The combined A+B rate is 4% (277 repos).

The average score is 40/100. Even popular, widely-used projects often lack basic health hygiene like security policies, CI configuration, and issue triage.

Security is the weakest dimension

Security has the lowest average at 30/100, while Documentation leads at 53/100. The 23-point gap between Documentation and Security reveals a systematic blind spot.

Security
30
Maintenance
33
DevOps
35
Testing
39
Architecture
40
Documentation
53

Language choice correlates with health

Repositories in systems languages and modern typed languages tend to score higher. This likely reflects engineering culture rather than language features.

Highest scoring

1. Rust (285) 61
2. OCaml (6) 57
3. TypeScript (856) 56
4. MDX (10) 54
5. Go (454) 53

Lowest scoring

1. MATLAB (7) 15
2. Objective-C (153) 18
3. TeX (20) 20
4. Less (5) 21
5. Pug (9) 24

Project type matters

5 project types across 7715 graded repos. Runtime scores highest at 70/100.

Runtime
70
Hybrid
59
Library
46
Application
40
Iac
37

Stars don't guarantee health

The top 20 most-starred repositories average 58/100 (overall avg: 40). Popularity doesn't automatically translate to good engineering practices.

build-your-own-x
28
486k stars
awesome
62
451k stars
freeCodeCamp
71
441k stars
public-apis
52
419k stars
free-programming-books
43
385k stars
developer-roadmap
53
352k stars
openclaw
91
346k stars
system-design-primer
25
341k stars

About this analysis

Scores are computed from publicly available repository metadata including CI configuration, security policies, documentation, test coverage indicators, dependency management, and issue triage. They reflect repository health practices, not code quality or project usefulness. Documentation-only and mirror repositories are excluded from grading. Data was collected on April 3, 2026.