What
that-depends has no AGENTS.md and no CLAUDE.md. It is the only repo in the org
without one, and it never had one: gh api repos/modern-python/that-depends/commits?path=CLAUDE.md
and the same for AGENTS.md both return empty. This is not fallout from the org-wide
rename in #51, which only touched repos that already had the file. It was never
onboarded.
It also has no CONTEXT.md and no docs/adr/.
Why it matters more here than anywhere else
that-depends is the org's most-used package and its only repo with real
contributor traffic. Across all PRs: 9 distinct authors, with xelandernt at 54 and
lesnik512 a minority at 28 of 100. Every other repo in the org is single-author
except fastapi-sqlalchemy-template.
So this is the one repo where an agent works alongside other people's changes, on the
codebase with the largest public surface, with zero steering.
Scope
Write an AGENTS.md describing the repo as it actually is, not a copy of
modern-di's conventions:
- It runs
mypy and pyrefly in addition to ruff and ty (lint-ci), which is
already noted in .github/AGENTS.md under Tooling.
- Its docs are on Read the Docs (
.readthedocs.yaml), not the org MkDocs setup.
- Its PR traffic is majority-external, so its workflow section should describe review
by a second party, which no other repo in the org has.
- It carries
context7.json and a .pre-commit-config.yaml, neither of which appear
elsewhere in the org.
Explicitly not in scope
- Its PR template stays exactly as it is.
.github/PULL_REQUEST_TEMPLATE.md is
the only template in the org measurably in use: xelandernt fills it in 8 of 9
PRs, checklist included. It is a local copy of the generic form differing only in
the mypy/pyrefly line, and that is a legitimate reason to stay local.
- Not the PR-body-as-spec convention. That convention assumes a single agent
author writing the body for its own benefit. This repo has human reviewers and
human PR authors, and the template that serves them already works.
CONTEXT.md and docs/adr/ are separate questions; do not create them speculatively.
Revisit trigger
None; this is a gap to close, not a decision to revisit.
What
that-dependshas noAGENTS.mdand noCLAUDE.md. It is the only repo in the orgwithout one, and it never had one:
gh api repos/modern-python/that-depends/commits?path=CLAUDE.mdand the same for
AGENTS.mdboth return empty. This is not fallout from the org-widerename in #51, which only touched repos that already had the file. It was never
onboarded.
It also has no
CONTEXT.mdand nodocs/adr/.Why it matters more here than anywhere else
that-dependsis the org's most-used package and its only repo with realcontributor traffic. Across all PRs: 9 distinct authors, with
xelanderntat 54 andlesnik512a minority at 28 of 100. Every other repo in the org is single-authorexcept
fastapi-sqlalchemy-template.So this is the one repo where an agent works alongside other people's changes, on the
codebase with the largest public surface, with zero steering.
Scope
Write an
AGENTS.mddescribing the repo as it actually is, not a copy ofmodern-di's conventions:mypyandpyreflyin addition toruffandty(lint-ci), which isalready noted in
.github/AGENTS.mdunder Tooling..readthedocs.yaml), not the org MkDocs setup.by a second party, which no other repo in the org has.
context7.jsonand a.pre-commit-config.yaml, neither of which appearelsewhere in the org.
Explicitly not in scope
.github/PULL_REQUEST_TEMPLATE.mdisthe only template in the org measurably in use:
xelanderntfills it in 8 of 9PRs, checklist included. It is a local copy of the generic form differing only in
the
mypy/pyreflyline, and that is a legitimate reason to stay local.author writing the body for its own benefit. This repo has human reviewers and
human PR authors, and the template that serves them already works.
CONTEXT.mdanddocs/adr/are separate questions; do not create them speculatively.Revisit trigger
None; this is a gap to close, not a decision to revisit.