Skip to content

GitHub standards

Every repository here says what it does not do. When a claim can be checked, the check ships with it. The ban on circumvention code in Molo is enforced by a test, so a patch that crosses that line fails CI instead of reaching anyone. The bridge between Claude Code and Codex is version gated: it refuses to write into a CLI version it has not been tested against.

That covers the code. This section covers the other half, the files around the code that tell someone arriving cold what a project is, what they are allowed to do with it, and how to send a change back.

Part of that half is measured by GitHub itself. Every public repository has a community profile: GitHub looks for a set of standard files and reports a percentage. The number comes from GitHub, I cannot edit it, and it is unflattering for most of what I publish. That is the reason it is on this page.

Synced on 4 August 2026, from GitHub’s community profile for github.com/nerln.

Repository Health README License CONTRIBUTING Code of conduct Issue template PR template
molo 100% yes yes yes yes no yes
claude-codex-bridge 71% yes yes yes no no no
argano 42% yes yes no no no no
beatrice-e-edoardo 42% yes yes no no no no
plancia 42% yes yes no no no no
rada 42% yes yes no no no no
scriba 42% yes yes no no no no
snapchat-memories-to-photos 42% yes yes no no no no
varo 42% yes yes no no no no
vesta 42% yes yes no no no no
wedding-invite-starter-kit 42% yes yes no no no no

Every repository in the list has a description and a README, and every one carries a license: MIT for most, GPL-3.0 for plancia, rada and varo.

Nine of eleven repositories sit at 42 percent. They all carry the same three things: a description, a README and a license. They are missing CONTRIBUTING, a code of conduct, an issue template and a pull request template.

Molo reaches 100 percent because it adds CONTRIBUTING, a code of conduct and a pull request template. It still has no issue template, and GitHub scores it 100 anyway. claude-codex-bridge reaches 71 percent by adding CONTRIBUTING and nothing else.

So the number tracks the presence of files. It does not read them. Nothing in the score asks whether the README describes the tool accurately, whether the tests pass, or whether the thing runs at all.

Contribution files answer questions that contributors ask. None of these repositories has a fork or an external contributor yet, and only one has an open issue. Writing a code of conduct for a project nobody has sent a patch to raises a number on a dashboard and changes nothing about the project.

The rule I use: a repository gets CONTRIBUTING when there is a real process to describe, meaning a test suite to run and a review path to follow. It gets templates when issues start arriving and I find myself asking the same questions twice. Molo went first because it has a hard boundary that a contributor could cross by accident, and that boundary needed writing down.

Whether the other nine follow will show up in this table, which is generated from the same API call for every repository. It moves when the repositories move.

Four things are not optional, and none of them appear in GitHub’s checklist.

A license on every repository, chosen before the first public commit.

A section in the README that says what the tool does not do, including the cases where the answer is a flat no.

A section that lists known limitations, including things that compile but have never been run on real hardware.

A check that would catch the claim if it were false, shipped in the repository, where you can run it.

The pages under Tools go through those one project at a time.