girlbossrush/authentik-agent-marketplace/testing
- community
Checking that the SKILL.md file is still in the repository.
Run authentik's tests: just the ones touching your change, a single test while iterating, the full backend suite, the Selenium e2e suite, or the web tests. Use when a contributor wants to run the Python or web tests, narrow a run to one test, watch an e2e run in a browser, or work out why a suite won't start. Linting and type checks live in linting; the environment the tests need in dev-environment.
Install
npx vskill@latest install girlbossrush/authentik-agent-marketplace/testingSKILL.md
authentik testing
Purpose
authentik has several test layers: Python unit and integration tests, a browser-driven e2e suite, and web frontend tests. Each has its own command and its own service prerequisites. This skill runs the right suite, narrows a run to a single test while iterating, and sorts out the setup a suite needs.
When to invoke
- "Run the Python e2e tests."
- "Run the backend unit tests" or "run just this one test."
- "Run the web / frontend tests."
- "The e2e suite won't start" or "tests pass in CI but fail locally."
Not this skill: linting and type checking (linting), or setting up the environment the tests run against (dev-environment, backend).
Common workflows
Steps are tagged by what you're doing: [shell] run a command, [verify] confirm the result. Commands run from the repo root unless noted.
Run a single backend test while iterating
Result: one test, class, or module runs fast against the kept test database.
- [shell]
uv run coverage run manage.py test tests.authentik.core.tests.test_models.ModelTest.test_user_settings --keepdb — drop the trailing .test_... to run the whole class, or pass a file path to run a module.
Gotchas: --keepdb reuses the test database (much faster on reruns); tests randomize order (pytest-randomly), so pass --randomly-seed=0 to make a debugging run deterministic. Verify: the focused test passes.
Run the full backend suite
Result: all backend tests with a coverage report.
- [shell]
make test(runs the suite with coverage; CI adds--parallel auto).
Verify: green, and htmlcov/index.html holds the coverage report.
Run the e2e (browser) suite, and watch it
Result: the Selenium-driven flow tests run, optionally visible in a browser.
- [shell] Start the e2e services:
docker compose -f tests/e2e/compose.yml up -d. - [shell] Run one:
uv run coverage run manage.py test tests.e2e.test_flows_login --keepdb
(or tests/e2e/ for all).
- [verify] Watch it live at
http://localhost:7900(passwordsecret).
Gotchas: e2e is slow because it drives a real browser. Run a specific test during dev, not the whole suite; the e2e services must be up first.
Run the web tests
Result: the frontend tests pass.
- [shell]
cd web/ && npm run test(Vitest unit tests);npm run test:e2e(Playwright).
Verify: green locally before pushing; CI's ci-web runs the same.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingplugins/developer/skills/testing/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/girlbossrush/authentik-agent-marketplace/testing)