cleveragents-core/features/definition_of_done.feature

Feature: Definition-of-Done gating
  As a plan executor
  I want apply to be blocked unless all DoD criteria are met
  So that only plans meeting completion criteria are committed

  Background:
    Given a definition of done test environment

  # -- DoD parsing ---------------------------------------------------------

  Scenario: Parse DoD text into criteria
    When I parse the DoD text "All tests pass\nNo lint errors\nCoverage above 90%"
    Then there should be 3 criteria parsed
    And criterion 0 text should be "All tests pass"
    And criterion 1 text should be "No lint errors"

  Scenario: Parse DoD with bullet points
    When I parse the DoD text "- Tests pass\n- No errors\n* Coverage ok"
    Then there should be 3 criteria parsed
    And criterion 0 text should be "Tests pass"

  Scenario: Parse DoD with numbered items
    When I parse the DoD text "1. Tests pass\n2. No errors"
    Then there should be 2 criteria parsed
    And criterion 0 text should be "Tests pass"
    And criterion 1 text should be "No errors"

  Scenario: Parse empty DoD returns no criteria
    When I parse an empty DoD text
    Then there should be 0 criteria parsed

  # -- Template rendering --------------------------------------------------

  Scenario: Render DoD template with plan arguments
    When I render the DoD template "Tests for {module} must pass" with argument "module" as "auth"
    Then the rendered DoD should be "Tests for auth must pass"

  Scenario: Render DoD template preserves unmatched placeholders
    When I render the DoD template "Deploy {service} to {env}" with argument "service" as "api"
    Then the rendered DoD should contain "{env}"

  # -- Text match evaluation -----------------------------------------------

  Scenario: TextMatchEvaluator passes when context contains criterion text
    Given DoD criteria from "All tests pass"
    And evaluation context with key "status" value "all tests pass"
    When I evaluate the criteria with TextMatchEvaluator
    Then all criteria should have passed

  Scenario: TextMatchEvaluator fails when context lacks criterion text
    Given DoD criteria from "All tests pass"
    And evaluation context with key "status" value "3 tests failed"
    When I evaluate the criteria with TextMatchEvaluator
    Then criterion 0 should have status "failed"

  Scenario: TextMatchEvaluator with regex pattern passes on match
    Given a DoD criterion with text "Coverage check" and pattern "coverage.*9[0-9]"
    And evaluation context with key "coverage" value "coverage=95%"
    When I evaluate the criteria with TextMatchEvaluator
    Then all criteria should have passed

  Scenario: TextMatchEvaluator with regex pattern fails on no match
    Given a DoD criterion with text "Coverage check" and pattern "coverage.*9[0-9]"
    And evaluation context with key "coverage" value "coverage=80%"
    When I evaluate the criteria with TextMatchEvaluator
    Then criterion 0 should have status "failed"

  Scenario: TextMatchEvaluator skips when context is empty
    Given DoD criteria from "All tests pass"
    And an empty evaluation context
    When I evaluate the criteria with TextMatchEvaluator
    Then criterion 0 should have status "skipped"

  Scenario: TextMatchEvaluator handles invalid regex gracefully
    Given a DoD criterion with text "Bad regex" and pattern "[invalid"
    And evaluation context with key "data" value "anything"
    When I evaluate the criteria with TextMatchEvaluator
    Then criterion 0 should have status "failed"
    And criterion 0 reasoning should contain "Invalid regex"

  # -- DoDSummary model ----------------------------------------------------

  Scenario: DoDSummary reports all_passed when no failures
    Given a DoDSummary with 2 passed and 0 failed
    Then the summary should report all passed
    And the DoD summary total should be 2

  Scenario: DoDSummary reports not all passed when failures exist
    Given a DoDSummary with 1 passed and 1 failed
    Then the summary should not report all passed
    And the summary failed count should be 1

  Scenario: DoDSummary empty check
    Given an empty DoDSummary
    Then the summary should be empty
    And the summary should not report all passed

  Scenario: DoDSummary to_cli_dict has expected fields
    Given a DoDSummary with 1 passed and 0 failed
    Then the summary CLI dict should have key "total"
    And the summary CLI dict should have key "all_passed"
    And the summary CLI dict should have key "criteria"

  Scenario: DoDSummary to_validation_summary is compatible
    Given a DoDSummary with 2 passed and 1 failed
    Then the validation summary should have required_passed 2
    And the validation summary should have required_failed 1
    And the validation summary should have dod_evaluated true

  # -- Evaluation context matching -----------------------------------------

  Scenario: TextMatchEvaluator matches criterion text in context keys
    Given DoD criteria from "lint_errors"
    And evaluation context with key "lint_errors" value "0"
    When I evaluate the criteria with TextMatchEvaluator
    Then all criteria should have passed

  Scenario: Multiple criteria with mixed results
    Given DoD criteria from "tests pass\nlint errors"
    And evaluation context with key "result" value "tests pass, no issues"
    When I evaluate the criteria with TextMatchEvaluator
    Then criterion 0 should have status "passed"
    And criterion 1 should have status "failed"

  # -- Coverage: edge cases ------------------------------------------------

  Scenario: Parse DoD with blank lines between criteria
    When I parse the DoD text "Tests pass\n\nNo errors"
    Then there should be 2 criteria parsed

  Scenario: Parse DoD with whitespace only line
    When I parse the DoD text "  \n- Tests pass"
    Then there should be 1 criteria parsed
    And criterion 0 text should be "Tests pass"

  Scenario: TextMatchEvaluator matches via individual key lookup
    Given DoD criteria from "all_tests_passed"
    And evaluation context with key "all_tests_passed" value "true"
    When I evaluate the criteria with TextMatchEvaluator
    Then all criteria should have passed

  # -- T1: all-SKIPPED summary gate behavior --------------------------------

  Scenario: DoDSummary all_passed is false when all criteria skipped
    Given a DoDSummary with all criteria skipped
    Then the summary should not report all passed
    And the summary skipped count should be 2

  # -- T2: bullet prefix yielding empty text --------------------------------

  Scenario: Parse DoD with bullet prefix but no text
    When I parse the DoD text "- \n- Tests pass"
    Then there should be 1 criteria parsed
    And criterion 0 text should be "Tests pass"

  # -- T5: multi-argument template rendering --------------------------------

  Scenario: Render DoD template with multiple arguments
    When I render the DoD template "{svc} in {env} OK" with arguments "svc"="api" and "env"="prod"
    Then the rendered DoD should be "api in prod OK"

  Scenario: Render DoD template prevents injection via values
    When I render the DoD template "Test {name}" with argument "name" as "{secret}"
    Then the rendered DoD should be "Test {secret}"

  # -- S3: length limit on template -----------------------------------------

  Scenario: Render DoD rejects oversized template
    Given a DoD template that is 11000 characters long
    When I render the oversized DoD template with argument "x" as "y"
    Then the DoD render should fail with "exceeds maximum"

  Scenario: Render DoD rejects oversized output
    Given a short DoD template with a huge argument value
    When I render the DoD template that produces oversized output
    Then the DoD render should fail with "exceeds maximum"

  # -- S2: ReDoS protection -------------------------------------------------

  Scenario: TextMatchEvaluator rejects oversized regex pattern
    Given a DoD criterion with text "long regex" and pattern that is 600 chars
    And evaluation context with key "data" value "anything"
    When I evaluate the criteria with TextMatchEvaluator
    Then criterion 0 should have status "failed"
    And criterion 0 reasoning should contain "exceeds maximum"