Coding Python

This skill WRITES implementation. Tests should already exist.

1. WRITE mode - Implementation doesn't exist or tests are failing

   • Check: Tests fail with ImportError or AssertionError
   • Action: Write implementation to make tests pass

2. FIX mode - Implementation exists but was rejected by reviewer

   • Check: Recent /auditing-python output shows REJECT with specific issues
   • Action: Read the rejection, fix the specific issues, re-run verification

Always check which mode before proceeding.

Before invoking this skill:

1. Tests must exist - Written by /testing-python
2. Tests must be reviewed - Approved by /auditing-python-tests
3. Spec must be loaded - Context from /spec-tree:contextualizing

If tests don't exist or aren't approved, go back to earlier steps.

WRITE Mode: Creating Implementation

Step 1: Understand Tests

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:

• What behaviors the tests verify
• What interfaces are expected (function signatures, classes)
• What the tests import (where implementation should live)

Step 2: Write Implementation (GREEN)

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

Step 3: Run Tests (Verify GREEN)

uv run --extra dev pytest {node_path}/tests/ -v

All tests should pass. If any fail, fix implementation and re-run.

Step 4: Refactor (Keep GREEN)

Clean up while keeping tests green:

1. Extract constants
2. Simplify
3. DRY

Step 5: Self-Verify

# 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.

FIX Mode: Fixing Rejected Implementation

Step 1: Read Rejection Feedback

Find the most recent /auditing-python output. Look for:

• Specific file:line locations
• Issue categories (magic values, missing DI, etc.)
• Required fixes

Step 2: Apply Fixes

For each rejection reason:

Step 3: Verify Fixes

# 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/

Step 4: Report What Was Fixed

## 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.

Mandatory Patterns

Named Constants

# ❌ 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

Dependency Injection

# ❌ 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

Type Annotations

# ✅ 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:

• All tests in {node}/tests/ pass
• Type checking passes (mypy)
• Linting passes (ruff check)
• Code uses named constants (no magic values)
• Code uses dependency injection (no direct external imports)
• All functions have type annotations
• All reviewer feedback addressed (if FIX mode)