next-evals-oss
AGENTS.md329 starsChanged 12 days agoMITView on GitHub
Read README.md first — it is the spec for this repo, not a summary of it. "Adding a new model", "Reasoning effort", "Model retention policy", and "Scoring and cost methodology" are the rules the board is published under. Skills live in .agents/skills/, with .claude/skills symlinked to that directory so Claude Code and other harnesses read the same copy. Almost every task here ends in numbers on a public board, and the work only counts once those numbers exist. An experiment…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,439 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository2 other files it gives its agents.
- add-eval-modelSkill
- release-eval-reportSkill
The file
We split this AGENTS.md into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 5
Read README.md first — it is the spec for this repo, not a summary of it.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- CLAUDE.md is a symlink to this file. Edit this one. -->Working in this repo
Read README.md first — it is the spec for this repo, not a summary of it. "Adding a new model", "Reasoning effort", "Model retention policy", and "Scoring and cost methodology" are the rules the board is published under.
Skills
Skills live in .agents/skills/, with .claude/skills symlinked to
that directory so Claude Code and other harnesses read the same copy.
| Skill | Use it for |
add-eval-model | Adding, registering, running, tiering, or publishing a model on nextjs.org/evals |
Finish the run
Almost every task here ends in numbers on a public board, and the work only counts once those numbers exist. An experiment with no results is not exported at all, so a PR that registers a model and stops changes nothing anyone can see — the README calls that state "a staging post, not a destination".
Merging to main here is what publishes: nextjs.org/evals fetches
agent-results.json from this repo's main on the server, and main CI invalidates
its cache. vercel/front used to carry a copy of the file and no longer does
(vercel/front#85415, 2026-09-09), so a results PR against the site repo is not part of
the task — only a site-side change that is genuinely about the site is. See
"Publishing to nextjs.org/evals" in the README.
Eval runs cost real money and take 30–90 minutes. That is the normal price of a task in this repo, not a reason to stop and ask. Run them, then report what they cost.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.