Agent-Skills.md

Agent Skills: Python Testing

Test Python services with pytest, async coverage, monkeypatch, and boundary-focused fakes. Use when writing Python tests, fixtures, async tests, regression tests, or dependency-isolated verification.

UncategorizedID: hoangnguyen0403/agent-skills-standard/python-testing

Repository

hoangnguyen0403/agent-skills-standard
HoangNguyen0403License: MIT
567164

Install this agent skill to your local

pnpm dlx add-skill https://github.com/HoangNguyen0403/agent-skills-standard/tree/HEAD/skills/python/python-testing

Skill Files

Browse the full folder contents for python-testing.

Download Skill

Loading file tree…

skills/python/python-testing/SKILL.md

Skill Metadata

Name
python-testing
Description
Test Python services with pytest, async coverage, monkeypatch, and boundary-focused fakes. Use when writing Python tests, fixtures, async tests, regression tests, or dependency-isolated verification.

Python Testing

Priority: P0 (CRITICAL)

Core Rule Anchors

  • [BE-TEST-01] Parametrized Tests for Equivalent Inputs: Prefer pytest.mark.parametrize for equivalent inputs and boundary permutations. Distinct scenarios may stay separate.
  • [BE-TEST-02] Ban on Brittle SQL String Matching: Test pure domain calculations directly or use fakes/testcontainers; do not regex-match SQL strings in mocks.
  • [BE-TEST-03] Ban on Shallow Assertions: Never assert only assert result without inspecting domain attributes and invariants.
  • [BE-TEST-04] Ban on Pass-Through Seam Mocks: Repositories and adapters must test contract compliance and error mapping; ban 1:1 pass-through mock echoing.
  • [BE-TEST-05] Bug-First Regression Lock: Every PR fixing a bug ticket or with title fix(...) must introduce a failing test reproducing the defect before fixing it.

Rules

  • Use pytest and pytest-asyncio for async boundaries.
  • Add regression tests around contract parsing, workflow routing, and verifier outcomes ([BE-TEST-05]).
  • Prefer pytest.mark.parametrize for equivalent inputs; distinct scenarios may stay separate ([BE-TEST-01]).
  • Prefer fakes or monkeypatch at external seams: DB, RPC, Telegram, subprocess, Docker ([BE-TEST-02], [BE-TEST-04]).
  • Verify behavior and domain invariants, not shallow truthiness or implementation trivia ([BE-TEST-03]).
  • Coverage is diagnostic and project-configured; verify risk-weighted critical paths rather than padding code for an arbitrary percentage.

Recipe

  1. Write the failing test first for the boundary or regression ([BE-TEST-05]).
  2. Patch one seam high enough to keep the test readable ([BE-TEST-04]).
  3. Assert structured outcomes: metadata, route result, blocker text, emitted command ([BE-TEST-03]).
  4. Use async tests for async code; avoid spinning real event loops manually.
  5. Cover negative paths for malformed packets, stale runtime state, and missing env.

Anti-Patterns

  • [BE-TEST-01] Stylistic parametrize rewrites: Do not demand parameterized rewrites of passing tests without a behavioral gap.
  • [BE-TEST-02] SQL regex matching: Do not regex-match SQL strings in mocks.
  • [BE-TEST-03] Shallow assertions: Never assert only assert result; always verify domain attributes and invariants.
  • [BE-TEST-04] Pass-through seam mocks: Ban 1:1 mock echoing without contract assertions.
  • [BE-TEST-05] Bug fix without reproduction test: Ban bug fixes without a reproduction test.
  • No live network or DB in unit tests: Use fakes or testcontainers.
  • No broad monkeypatch spray: Patch the boundary, not every helper under it.
  • No smoke-only proof for shared runtime changes: Add focused assertions too.

References