apache-airflow
Route Apache Airflow repo tasks across Dag authoring, operations, providers, deployment, and contribution workflows.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- apache-airflow
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Route Apache Airflow repo tasks across Dag authoring, operations, providers, deployment, and contribution workflows.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/repo-provenance.md2.4 KB
- 参考文档references/repo-routing-metadata.json316 B
- 参考文档references/troubleshooting.md3.2 KB
- 脚本scripts/check_airflow_skill_environment.py3.8 KB
03 SKILL.md 正文
<!-- SPDX-License-Identifier: Apache-2.0 -->
Apache Airflow Repo Skill
Use this skill when a task names Apache Airflow, the apache-airflow Python package, the Airflow monorepo, Airflow Dags, Task SDK, airflow/airflowctl, providers, official Docker images, Helm chart, Breeze, or Airflow contribution rules.
This skill is a router. Read the nearest sub-skill for workflow depth and use repo-level references only for shared context.
Start Here
- Read
references/repo-provenance.mdbefore relying on this skill for a checkout; refresh the skill if the commit, dirty state, or public package versions no longer match. - Read
references/troubleshooting.mdfor cross-cutting installation, import, routing, and validation failures. - Use
scripts/check_airflow_skill_environment.pyfor a quick installed-package and helper-script check when a Python environment is available. - Choose exactly one primary sub-skill from the route map, then follow its linked references and bundled scripts.
Route Map
sub-skills/authoring-task-sdk/— write, migrate, validate, or debug Airflow 3 Dag authoring code usingairflow.sdk, Task SDK, TaskFlow, dynamic task mapping, assets, timetables, Params, XCom/context, and standard provider operators/sensors.sub-skills/operations-cli-api/— install or run Airflow, inspect configuration, useairfloworairflowctl, choose Stable REST API vs CLI, operate core components, test/backfill Dags from the command line, and troubleshoot metadata DB/API/server state.sub-skills/providers-extensions/— use provider packages, standard operators/sensors/hooks, custom operators/hooks/sensors, plugins/listeners/timetables/notifiers/extra links, provider metadata, and provider package conventions.sub-skills/deployment-helm-docker/— plan or debug deployments with the official Helm chart and Docker images, including chart values, custom images, Dag delivery, logs, secrets/config, migrations, and autoscaling.sub-skills/contribution-tooling/— change the Airflow repository safely with Breeze,uv,prek, selective checks, docs/news/changelog rules, generated-file constraints, PR conventions, and component-specific tests.
Common Routing Decisions
- If the task is about a user Dag file, start with
authoring-task-sdkeven if the symptom appears throughairflow dags test; return tooperations-cli-apionly for command/config execution details. - If a provider import fails in a Dag, use
providers-extensionsfor package/extras/provider metadata andauthoring-task-sdkfor Dag structure. - If a scheduler or Dag processor cannot parse Dags in Kubernetes, use
deployment-helm-dockerfor image/Dag delivery andoperations-cli-apifor component/database diagnosis. - If the task is to edit Airflow source code, use
contribution-toolingfirst, then route to the domain sub-skill for product behavior. - Java SDK, Go SDK, new language SDK, translation, and provider release-manager workflows have specialized repo-local skills in the source checkout; this generated skill keeps only routing-level notes for those areas.
Install and Import Facts
- Public package names verified for this snapshot:
apache-airflow,apache-airflow-core,apache-airflow-task-sdk,apache-airflow-ctl, andapache-airflow-providers-standard. - Public import roots verified for this snapshot:
airflow,airflow.sdk,airflowctl, andairflow.providers.standard. - Airflow installation should use official constraints for repeatability; do not recommend broad
[all]extras unless the user explicitly needs broad provider coverage. - Dag authors should use the Airflow 3 public interface from
airflow.sdk; avoid internal metadata DB access from task code.
Bundled Helpers
scripts/check_airflow_skill_environment.pychecks installed distribution metadata/imports and confirms bundled helper scripts are present.- Sub-skill helpers provide targeted checks: Dag file parsing, CLI parser inspection, provider metadata sanity checks, Helm values summaries, and first-pass Airflow contribution command recommendations.
Safety Boundaries
- Runtime instructions in this skill are self-contained and do not depend on the original repository checkout.
- Do not copy local environment paths, private prefixes, or machine-specific setup details into user-facing work.
- Do not run Airflow native tests, examples, Docker builds, Helm cluster commands, or release-management scripts unless the active sub-skill classifies them as safe for the current environment.
- In Airflow prose, write Dag in title case; preserve literal code/config/CLI tokens such as
DAG,dag_id,dag,airflow dags list, andget_dagexactly.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).