pyfixest
AGENTS.md363 starsChanged 12 days agoMITView on GitHub
pyfixest brings R's fixest workflow to Python: fixed-effects estimation, formula syntax, inference, and post-estimation, with Rust kernels for hot loops. - Mirror fixest in overlapping user-facing behavior unless a difference is intentional; compatibility policy owns those decisions. Never silently fit a different model or return numbers for an unsupported path. - Start with the nearest implementation and its tests. Complete the request in small, self-contained changes that are easy to review; see PR sizing. Keep independent fixes separate, including numerical…
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 31,700 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 repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- change-verificationSkill
- implementation-strategySkill
- pr-handoffSkill
- pr-reviewSkill
The file
We split this AGENTS.md into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 12
Start with the nearest implementation and its tests.
- Distinctive: line 15
Keep independent fixes separate, including numerical corrections discovered during an invariant refactor.
- Distinctive: line 17
Prefer code a maintainer can follow from inputs to results.
- Very specific: line 20
Use pixi run for Python, pytest, lint, docs, and R commands.
- Very specific: line 50
Never hand-edit generated docs/reference/, docs/_freeze/, docs/_site/, .coverage, or coverage.xml.
- Distinctive: line 52
Follow Git approval policy: never commit to master, rewrite history without approval for that specific rewrite, or merge agent-authored …
≈ “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
pyfixest — guide for coding agents
pyfixest brings R's fixest workflow to Python: fixed-effects estimation,
formula syntax, inference, and post-estimation, with Rust kernels for hot loops.
Essentials
- Mirror
fixestin overlapping user-facing behavior unless a difference is intentional; compatibility policy owns those decisions. Never silently fit a different model or return numbers for an unsupported path. - Start with the nearest implementation and its tests. Complete the request in small, self-contained changes that are easy to review; see PR sizing. Keep independent fixes separate, including numerical corrections discovered during an invariant refactor.
- Prefer code a maintainer can follow from inputs to results. An abstraction earns its place by clarifying an operation, owning an invariant, or serving real callers—not by adding another layer to pass through.
- Use
pixi runfor Python, pytest, lint, docs, and R commands. Usepixi task listto discover tasks. Select checks from the testing policy, including before/after benchmarks for performance-relevant estimation changes.
Load only the workflow the task needs
| Task | Read |
| Implement estimator, inference, formula, post-estimation, public API, or shared-core changes | implementation-strategy |
| Select or report checks for a change | change-verification |
| Review a diff, self-review, or walk through changed code together | pr-review, choosing the requested mode |
| Prepare commits, push, open a PR, or curate history | pr-handoff |
| Create or edit a GitHub issue | issue style |
A question or prose correction does not require an implementation plan. Reuse decisions and checks already established unless the change invalidates them. Internal workflow steps need not appear as checklists in replies or PRs.
docs/developer/ owns detailed policy; skills apply it. Read the relevant
sections, not every linked document. Architecture
owns code conventions and extension seams; Git and PR style
owns scope and publication. These documents are not part of the rendered site.
CLAUDE.md imports this file; .claude/skills/ links to the shared skills.
docs/skills.md is guidance for package users, not contributors.
Boundaries
- Preserve unrelated user changes. Change
pixi.lockorCargo.lockonly for intentional dependency work; avoid unrelated formatting. - Never hand-edit generated
docs/reference/**,docs/_freeze/**,docs/_site/**,.coverage, orcoverage.xml. - Follow Git approval policy:
never commit to
master, rewrite history without approval for that specific rewrite, or merge agent-authored work. - Edit contributor guidance only when the task concerns it. Otherwise propose a discovered policy gap at handoff instead of adding rules during a code fix.
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.