Write Python code that's tested and type-safe. Use when coding Python or implementing features.
This skill WRITES implementation. Tests should already exist.
WRITE mode - Implementation doesn't exist or tests are failing
FIX mode - Implementation exists but was rejected by reviewer
/auditing-python output shows REJECT with specific issuesAlways check which mode before proceeding.
Before invoking this skill:
/testing-python/auditing-python-tests/spec-tree:contextualizingIf tests don't exist or aren't approved, go back to earlier steps.
Read the existing tests to understand:
# Read test files
cat {node_path}/tests/*.py
# Run tests to see failures
uv run --extra dev pytest {node_path}/tests/ -v
Understand:
Write minimal code that makes tests pass.
Code standards (per /standardizing-python):
# ✅ Type annotations on ALL functions
def process_order(order: Order, config: Config) -> OrderResult: ...
# ✅ Named constants for all literals
MIN_ORDER_VALUE = 10
MAX_ITEMS = 100
# ✅ Dependency injection for external dependencies
@dataclass
class Deps:
run_command: CommandRunner
uv run --extra dev pytest {node_path}/tests/ -v
All tests should pass. If any fail, fix implementation and re-run.
Clean up while keeping tests green:
# Type checking
uv run --extra dev mypy src/
# Linting
uv run --extra dev ruff check src/
# Tests one more time
uv run --extra dev pytest {node_path}/tests/ -v
All must pass before declaring complete.
Find the most recent /auditing-python output. Look for:
For each rejection reason:
| Rejection Category | Fix Action |
|---|---|
| Magic values | Extract to named constants |
| Missing type annotations | Add types to all functions |
| Direct external imports | Refactor to dependency injection |
| Deep relative imports | Change to absolute imports |
Missing -> None |
Add return type |
| Security issues | Fix the vulnerability (don't suppress) |
# Run tests
uv run --extra dev pytest {node_path}/tests/ -v
# Type checking
uv run --extra dev mypy src/
# Linting
uv run --extra dev ruff check src/
## Implementation Fixed
### Issues Addressed
| Issue | Location | Fix Applied |
| ----------- | --------------- | --------------------------------- |
| Magic value | handler.py:45 | Extracted to MAX_RETRIES constant |
| Missing DI | processor.py:12 | Added ProcessorDeps dataclass |
### Verification
All tests pass. Types and lint clean. Ready for re-review.
# ❌ REJECTED
def validate_score(score: int) -> bool:
return 0 <= score <= 100
# ✅ REQUIRED
MIN_SCORE = 0
MAX_SCORE = 100
def validate_score(score: int) -> bool:
return MIN_SCORE <= score <= MAX_SCORE
# ❌ REJECTED
import subprocess
def sync_files(src: str, dest: str) -> bool:
result = subprocess.run(["rsync", src, dest])
return result.returncode == 0
# ✅ REQUIRED
@dataclass
class SyncDeps:
run_command: CommandRunner
def sync_files(src: str, dest: str, deps: SyncDeps) -> bool:
returncode, _, _ = deps.run_command.run(["rsync", src, dest])
return returncode == 0
# ✅ All functions have full type annotations
def get_user(user_id: int) -> User | None:
users: list[User] = fetch_users()
return next((u for u in users if u.id == user_id), None)
WRITE mode output:
## Implementation Complete
### Node: {node_path}
### Files Created/Modified
| File | Action | Description |
| ---------------- | ------- | ------------- |
| `src/handler.py` | Created | Order handler |
### Verification
- Tests: ✓ Pass
- Types: ✓ Pass
- Lint: ✓ Pass
Ready for review.
FIX mode output:
## Implementation Fixed
### Issues Addressed
| Issue | Location | Fix Applied |
| ------- | ----------- | ----------- |
| {issue} | {file:line} | {fix} |
### Verification
All checks pass. Ready for re-review.
Task is complete when:
{node}/tests/ passmypy)ruff check)