菜单
Reviews code changes in azure-ai-ml package for quality, Azure SDK compliance, and best practices. Use when reviewing code, checking pull requests, or when user asks to review changes or check code quality in azure-ai-ml.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/Azure/azure-sdk-for-python --skill do-code-review命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | do-code-review |
| description | Reviews code changes in azure-ai-ml package for quality, Azure SDK compliance, and best practices. Use when reviewing code, checking pull requests, or when user asks to review changes or check code quality in azure-ai-ml. |
Reviews uncommitted changes (staged and unstaged files) in the azure-ai-ml package, focusing on Azure SDK Python design guidelines, type safety, testing patterns, and API consistency.
Unless otherwise specified, review all uncommitted changes in the current branch (staged and unstaged files) within sdk/ml/azure-ai-ml/. This includes new files, modified files, and any pending changes that haven't been committed yet.
create_or_update not create_or_replace)def get_job(name, subscription_id) should be def get_job(name, **kwargs)Key Patterns:
begin_* for LROs, list_* for paginatorsOptional, Union, TYPE_CHECKINGAny without justification, bare dict/listdef process_data(data) should be def process_data(data: Dict[str, Any]) -> ProcessedDataCommon Fixes:
# Bad
def get_config(name):
return config
# Good
def get_config(name: str) -> Optional[Dict[str, Any]]:
return config
Watch for:
client-method-should-not-use-static-methodmissing-client-constructor-parameter-credentialclient-method-has-more-than-5-positional-arguments_async modulesasync with for clients, awaiting coroutines correctlyawait, sync code in async modulesself._client.get() in async should be await self._client.get()Pattern:
# In azure/ai/ml/aio/operations/
async def create_or_update(
self,
entity: Job,
**kwargs: Any
) -> Job:
async with self._lock:
result = await self._service_client.create_or_update(...)
return result
HttpResponseError, ResourceNotFoundError, proper validationexcept:, catching Exception without re-raising, missing validationPattern:
from azure.core.exceptions import ResourceNotFoundError, HttpResponseError
try:
result = self._operation.get(name)
except ResourceNotFoundError:
raise ResourceNotFoundError(f"Job '{name}' not found")
except HttpResponseError as e:
raise HttpResponseError(f"Failed to retrieve job: {e.message}")
Structure:
azure/ai/ml/
├── operations/ # Sync operations
│ ├── job_operations.py
│ └── model_operations.py
└── aio/operations/ # Async operations (mirror structure)
├── job_operations.py
└── model_operations.py
Entity Pattern:
@dataclass
class Job(Resource):
"""Job entity."""
name: str
experiment_name: Optional[str] = None
def _to_rest_object(self) -> RestJob:
"""Convert to REST representation."""
...
@classmethod
def _from_rest_object(cls, obj: RestJob) -> "Job":
"""Create from REST representation."""
...
pytest, proper test isolation, fixture usageTest Structure:
class TestJobOperations:
"""Test job operations."""
def test_create_job(self, client: MLClient, mock_workspace: Mock) -> None:
"""Test job creation."""
job = Job(name="test-job")
result = client.jobs.create_or_update(job)
assert result.name == "test-job"
@pytest.mark.recorded
def test_get_job_recorded(self, client: MLClient) -> None:
"""Test getting job with recording."""
...
Docstring Pattern:
def create_or_update(
self,
job: Job,
**kwargs: Any
) -> Job:
"""Create or update a job.
:param job: The job entity to create or update.
:type job: ~azure.ai.ml.entities.Job
:keyword bool skip_validation: Skip validation of the job.
:return: The created or updated job.
:rtype: ~azure.ai.ml.entities.Job
:raises ~azure.core.exceptions.HttpResponseError: If the request fails.
.. admonition:: Example:
.. code-block:: python
from azure.ai.ml.entities import Job
job = Job(name="my-job")
result = ml_client.jobs.create_or_update(job)
"""
Deprecation Pattern:
import warnings
def old_method(self, param: str) -> None:
"""Deprecated method.
.. deprecated:: 1.2.0
Use :meth:`new_method` instead.
"""
warnings.warn(
"old_method is deprecated, use new_method instead",
DeprecationWarning,
stacklevel=2
)
self.new_method(param)
TokenCredential, proper token refresh, sanitized loggingPattern:
from azure.core.credentials import TokenCredential
class MLClient:
def __init__(
self,
credential: TokenCredential,
subscription_id: str,
**kwargs: Any
):
self._credential = credential # Store, don't log
# Never log credential or tokens
Pagination Pattern:
def list(self, **kwargs: Any) -> Iterable[Job]:
"""List jobs with pagination.
:return: An iterable of jobs.
:rtype: ~azure.core.paging.ItemPaged[~azure.ai.ml.entities.Job]
"""
return self._operation.list(...) # Returns ItemPaged
sdk/ml/azure-ai-ml/Organize findings by priority and category:
Good patterns worth highlighting
For each issue:
Focus on issues that impact SDK quality, user experience, backwards compatibility, and Azure SDK guideline compliance.
EXPERIMENTAL. Create and test a classify-and-route Azure AI Content Understanding pipeline for packets that contain multiple document types (e.g. invoice + bank statement + loan application in one PDF). Walks per-type schema authoring → outer classifier wiring → batch test → category-aware stdout summary using the typed ContentUnderstandingClient. Use when the user has mixed-document packets.
EXPERIMENTAL. Iteratively author and test a custom Azure AI Content Understanding analyzer for a folder of **document** files (PDFs, page images) of a single type. Walks layout extraction → schema authoring → validation → batch test → agent review → refine cycle using the typed ContentUnderstandingClient. Use when the user wants to author a custom analyzer for invoices, contracts, forms, or any other single-type document set.
Run a specific sample for the Azure AI Content Understanding SDK. Use when users want to run a particular sample like sample_analyze_url.py or sample_analyze_invoice.py.
基于 SOC 职业分类
