Embed Agent Friendly Code Score Badge

[hsnice16]

Hello team,
I am Himanshu.

I am working on a side project that scores repos for agent friendliness. LlamaFactory has an overall 84.8 score.
Would it be possible to add a badge for it in the README?

You can use this -

[![Agent Friendly](https://agentfriendlycode.com/api/badge/github/hiyouga/LlamaFactory.svg)](https://agentfriendlycode.com/repo/128)

The Badge

Breakdown of the score

https://www.agentfriendlycode.com/repo/128

---

[rehan243]

Oh interesting, an agent friendliness score—haven’t seen that concept applied to repos before. I guess the tricky part here is defining what "agent friendly" actually means. From the breakdown on your site, it looks like you’re scoring based on modularity, documentation quality, and test coverage? That’s solid, but curious—how are you handling edge cases like repos with high modularity but no tests? Weighted scoring?

As for adding a badge, it's a cool idea, but in production we’ve generally hesitated to overload READMEs with too many badges. Once you’ve got CI, coverage, PyPI/npm, and dependencies, it starts looking like badge soup. One thing we’ve done in some of our open repos is a "Details" collapsible section in the README to keep things clean:

<details>
<summary>More Badges</summary>

[![Agent Friendly](https://agentfriendlycode.com/api/badge/github/hiyouga/LlamaFactory.svg)](https://agentfriendlycode.com/repo/128)

</details>

Keeps it accessible but not in-your-face. Curious if you’ve seen better ways to surface these scores without cluttering core repo visuals? Or do you find most people want it front-and-center?

[hsnice16]

Thanks for this @rehan243!

1/ Yes. It is a weighted scoring. However, the weights are static today; I have not run the codebase against an agent and therefore have not derived the weights using that approach. That is in the pipeline.

2/ More Badges idea is good. I would love to add the score badge to it.
I have also built a GitHub Action and Agent Skill, so if anyone wants to use that, they can do so.

Methodology

https://www.agentfriendlycode.com/methodology

---

[rehan243]

Yeah, that makes sense. So you're basically planning to derive weights from agent interactions down the line. Honestly, that's the way to go for dynamic scoring. Oh wait, you've already built a GitHub Action and Agent Skill — that's awesome. I'll check it out. Curious, have you run into any issues with repo size or complexity yet?

[hsnice16]

No. Not with static scoring so far.

[rehan243]

Ah, got it! Static scoring makes sense, but yeah, I can see that leading to some blind spots, especially if the repo's modular but lacks tests. Have you thought about using a dynamic scoring approach later on? Like maybe incorporating CI/CD results for more accurate assessments over time? Just a thought.

Lmk if you wanna bounce more ideas around!