Pre-Commit Check

Run proportional local checks for the files being changed, catching relevant issues before pushing without defaulting to the full test suite.

When to Use

• Before creating a commit or PR
• User says "check my changes", "pre-commit", "verify before commit", "ready to push"
• After making any code changes to .py files

Steps

1. Identify Changed Files

git diff --name-only HEAD
git diff --name-only --cached
git status --short

Collect all changed/added .py files.

Classify the change by affected area: workflow, docs, agent skill, packaged configuration, isolated Python module, package-wide behavior, or cross-cutting infrastructure.

2. Run Black Formatting Check

This is the first CI gate and will cause immediate failure:

black --check --diff --color ./

If it fails, run black . and review the formatting changes.

3. Check Public API Documentation Coverage

Run the same read-only gate used by CI:

python docs/scripts/check_api_docs.py

If it reports missing exports, use $update-api-docs to add useful Sphinx entries and descriptions. Do not change __all__ solely to make this check pass.

4. Check Apache 2.0 Copyright Header

Every .py file must begin with the 15-line copyright block. For each changed/new .py file, verify the first line is:

# ----------------------------------------------------------------------------

The full header template:

# ----------------------------------------------------------------------------
# Copyright (c) 2021-2026 DexForce Technology Co., Ltd.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ----------------------------------------------------------------------------

5. Check from __future__ import annotations

Every .py file must have this import (after the header, before other imports). This enables A | B syntax and forward references.

6. Check __all__ in Public Modules

For any new or modified module under embodichain/, verify it defines __all__ listing all public symbols. Example:

__all__ = ["MyClass", "my_function"]

Skip this check for __init__.py files that only re-export via from . import *.

7. Check Docstrings on Public APIs

For any new public function, class, or method:

• Must have a Google-style docstring
• Must include Args: section if it takes parameters
• Must include Returns: section if it returns a value
• Use .. attention:: or .. tip:: directives for non-obvious behavior

8. Check Type Annotations

For any new public API:

• All parameters must have type hints
• Return type must be annotated
• Use A | B over Union[A, B]
• Use TYPE_CHECKING guard for imports that would cause circular dependencies

9. Check @configclass Usage

For any new configuration class:

• Must use @configclass decorator (not bare @dataclass)
• Must use from dataclasses import MISSING for required fields
• Import from embodichain.utils import configclass

10. Select and Run Relevant Tests

Do not treat the CI test job as a requirement to run pytest tests locally for every change. Choose the smallest command set that exercises the affected behavior:

Skip runtime tests when no executable behavior is affected, but still run the appropriate syntax or configuration validator. Run the full suite only when:

• the change affects shared core behavior used throughout the repository;
• global dependencies, test configuration, or environment initialization changed;
• several subsystems are modified together;
• a release-critical behavior cannot be validated narrowly; or
• the user explicitly requests it.

For changes identified as performance-sensitive during task analysis, follow implement-performance and select focused correctness checks for the algorithm, batch, memory or concurrency contracts being changed. Validate material performance claims using benchmark, comparing equivalent workloads and quality settings, including relevant small-batch behavior and peak memory. Report measured gains, tradeoffs and regressions with hardware and timing scope. If required hardware is unavailable, state what remains unvalidated; passing CPU tests does not establish GPU performance. Ordinary edits do not require a benchmark run.

Before starting a command likely to take more than two minutes, report the selected scope and why narrower validation is insufficient. Honor explicit user instructions to skip or narrow tests.

For canonical skill directories, use:

python "${CODEX_HOME:-$HOME/.codex}/skills/.system/skill-creator/scripts/quick_validate.py" \
  .agents/skills/<skill>

If the personal skill installation differs, locate the available skill-creator validator through the session skill catalog instead of assuming a username.

Thin .claude/skills/ and .github/copilot/ adapters should point back to the canonical .agents/skills/<skill>/SKILL.md; do not duplicate the full instructions in adapters.

11. Check Test Coverage and Necessity

Apply the shared test necessity review to the changed behavior and any added or extended cases. Confirm that observable behavior and real regressions have meaningful coverage; reuse sufficient existing tests without requiring a new test file for every public module or function.

Report an actual unverified contract or regression when coverage is missing. Review additions for redundancy and verification value, and briefly identify the existing coverage relied upon or why runtime tests are unnecessary. Check test file headers with the other changed Python files in step 4.

12. Summary Report

Output a pass/fail summary:

Pre-Commit Check Results
========================
[PASS] Black formatting
[PASS] Public API docs coverage
[PASS] Apache 2.0 headers (5/5 files)
[FAIL] from __future__ import annotations — missing in: foo.py
[PASS] __all__ exports
[PASS] Docstrings on public APIs
[PASS] Type annotations
[PASS] @configclass usage
[PASS] Targeted tests — tests/foo/test_bar.py
[N/A] Full test suite — isolated change covered by targeted tests
[PASS] Coverage and test necessity — existing coverage reused; no redundant additions

Fix the above issues before committing.

What CI Checks

The project's CI pipeline (.github/workflows/main.yml) runs:

1. lint job: black --check --diff --color ./
2. lint job: python docs/scripts/check_api_docs.py
3. test job: proportional pytest groups after lint passes
4. build job: Sphinx docs build after lint passes

This skill always covers the relevant lint and structural checks, then selects tests proportionally. It does not require reproducing the entire CI pipeline for every local change.

Common Mistakes

Quick Reference