# Gherkin

> Write clear, idiomatic formal Gherkin feature specifications. Use when creating or editing .feature files, Scenario Outlines, Examples tables, Background sections, dialect keywords, or step wording. Do not use for behavior clarification before a formal artifact; use behavior-driven-development.

- Skill: `nledford/gherkin` (Agent Skill)
- Install (CLI): `npx skillmds@latest add nledford/gherkin`
- Raw SKILL.md: https://api.skillmd.com/api/skills/nledford/gherkin/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: nledford (https://skillmd.com/u/nledford)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/nledford/gherkin

---


# Gherkin

Use Gherkin to write business-readable behavior specifications that can be
reviewed by non-engineers and connected to automated tests. Use the official
Gherkin reference as the conceptual baseline:
https://cucumber.io/docs/gherkin/reference

When the behavior, rule, or acceptance example is unclear, load
[`behavior-driven-development`](../behavior-driven-development/SKILL.md) before
writing formal syntax. Do not make a `.feature` file the default output for
informal behavior clarification.

For formal photo/video DAM scenarios, compose with
[`digital-asset-management`](../digital-asset-management/SKILL.md) to establish
the asset lifecycle vocabulary and invariants before writing feature syntax.

## Inspect Local Conventions First

Before editing, inspect existing `.feature` files for dialect declarations,
keyword style, tags, indentation, and naming. Identify the configured runner,
its narrowest feature/scenario command, and the corresponding step definitions
or glue code; align new steps with the suite's existing vocabulary.

## Core Structure

```gherkin
Feature: Short description of the capability

  Rule: Optional business rule that groups related scenarios

    Background:
      Given shared context for every scenario in this rule

    Scenario: Specific behavior title
      Given relevant context
      When the actor performs the meaningful action
      Then an observable outcome occurs
```

Use only the structure needed for clarity. A small feature may need only
`Feature` and a few `Scenario` blocks.

## Keyword Guidance

- `Feature`: name the capability or business goal, not a component.
- `Rule`: group scenarios that demonstrate one business rule.
- `Scenario`: use for one concrete example.
- `Scenario Outline`: use when the same behavior must be exercised with several
  meaningful data variations.
- `Examples`: keep tables small and focused on the variables that change the
  outcome.
- `Background`: use only for shared context that every scenario in the feature or
  rule genuinely needs.
- `Given`: establish relevant preconditions or state.
- `When`: describe the single meaningful action or event.
- `Then`: describe observable outcomes.
- `And` / `But`: continue the previous step type when it improves readability;
  do not use them to hide multiple unrelated behaviors.

## Writing Good Steps

Good steps are declarative:

```gherkin
Given Avery has an active workspace membership
When Avery uploads a duplicate photo
Then the upload is rejected as a duplicate
And the original photo remains unchanged
```

Avoid UI scripts and implementation details:

```gherkin
Given the user is on "/photos/new"
When they click "Choose File" and click "Upload"
Then the API returns 409
And the duplicate_photos table has 1 row
```

Use UI, HTTP, or database details only when they are the behavior contract being
specified.

## Scenario vs Scenario Outline

Use `Scenario` when one concrete example communicates the behavior clearly.

Use `Scenario Outline` when:

- The same rule has several important input/output pairs.
- The examples table is shorter and clearer than repeated scenarios.
- Each row protects a distinct case that should fail independently.

Do not use `Scenario Outline` just to compress unrelated behaviors into one
table. Split scenarios when the narrative, preconditions, or expected outcomes
change meaningfully.

## Background Discipline

Use `Background` sparingly:

- Keep it short, usually one to four steps.
- Include only context that every scenario needs.
- Avoid important assertions in `Background`; scenarios should contain their own
  outcomes.
- Prefer explicit scenario setup when shared setup makes the scenario hard to
  understand.

## Readability Rules

- Write concise scenario titles that finish the sentence "Scenario: ...".
- Keep feature files readable by product, QA, support, and domain experts.
- Prefer domain terms over technical terms.
- Avoid duplicate scenarios that assert the same rule with unimportant data
  changes.
- Keep steps stable across UI redesigns and implementation refactors.
- Align step wording with existing step definitions when editing an established
  suite, but do not preserve misleading wording if it obscures behavior.

## Validation Checklist

- Run the narrowest available Gherkin parser, feature/scenario command, or runner
  lane after editing; start with the affected file or scenario.
- If no parser or executable lane is available, validate syntax and local
  conventions manually and report that automated validation was unavailable or
  skipped, with the reason.
- The file uses valid Gherkin keywords and indentation.
- Every scenario has at least one observable `Then`.
- Each `When` describes the behavior-triggering action or event.
- Examples tables have headers and values for every parameter.
- No scenario depends on order from another scenario.
- Tags, if used, describe execution needs or meaningful categories rather than
  temporary implementation notes.

## Language Usage

- **Python:** map feature steps to pytest-bdd, behave, or project-specific
  adapters only when the repository already uses them or the user explicitly asks
  for executable `.feature` files. Keep step definitions thin and push domain
  rules into Python modules with ordinary tests.
- **Rust:** use Cucumber-style crates or custom feature runners only when the
  repository already has that lane. Otherwise keep Gherkin as acceptance criteria
  and implement behavior with Rust unit/integration tests.
- **JavaScript/TypeScript and other languages:** connect `.feature` files to the
  existing test runner only when it improves stakeholder-readable coverage. Do
  not introduce a Gherkin runner for simple changes where clear test names or
  inline Given/When/Then notes would be enough.