"Maintain the Flower repository itself: tests, public API exposure,
Scanned 9/8/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill repository-maintenance --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Repository Maintenance?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-repository-maintenance-arex-skill)More formats (shields.io, HTML) on the badges page.
---
name: repository-maintenance
description: "Maintain the Flower repository itself: tests, public API exposure,
protobufs, Alembic migrations, docs, packaging, and contributor guardrails."
metadata:
disco-role: operating
disable-model-invocation: true
license: Apache 2.0
---
# Flower Repository Maintenance
Use this sub-skill when the task is about changing, reviewing, or validating the
Flower repository itself. It is not the right route for ordinary Flower App,
strategy, dataset, simulation, or deployment usage unless the user is asking how
the repository should test or maintain those surfaces.
## Route first
- For contributor command selection, PR review expectations, dev environment
setup, lint/type/test families, docs builds, package builds, protobufs, and
Alembic migrations, read [references/maintenance.md](references/maintenance.md).
- For failures involving stale generated protobufs, migration drift, missing docs
prerequisites, public API exposure, or using a runtime environment for dev
checks, read [references/troubleshooting.md](references/troubleshooting.md).
- For a safe importability check of public `__all__` exports in an installed or
editable package, run the bundled
[scripts/check_public_api.py](scripts/check_public_api.py)
from this sub-skill directory in an environment where `flwr` or
`flwr_datasets` is importable.
## Maintainer defaults
- Apply the repository PR review rules: prefer necessary, simple, readable,
locally consistent changes; flag PRs that mix unrelated concerns.
- Use package-local commands. Framework commands normally run from `framework/`;
Flower Datasets commands normally run from `datasets/`.
- For framework work, prefer `uv run --no-sync --python=3.11.14 <command>` after
the dev environment has been intentionally synchronized.
- Do not edit generated protobuf files directly; edit `.proto` sources and
regenerate outputs.
- Do not hand-write a normal Alembic schema-diff migration; use Flower's
generator first, then review the generated revision.
- Public Python API exposure is compatibility-sensitive. Follow `__all__` from
the root package, add import-path tests, and avoid exposing implementation
modules accidentally.
## Exclusions
- Do not run broad e2e scripts casually. They can modify app-local config,
generate certs/databases, start background services, or require ports.
- Do not treat the minimum runtime/editable install used for package inspection
as a complete contributor environment; dev checks require dev groups and, for
docs, system prerequisites.
- Do not commit generated/cache outputs such as docs builds, wheel artifacts,
caches, or local virtual environments.

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!