x-plan-tests

Global Output Policy

• Language: English ONLY.
• Tone: Technical, Direct, and Concise.
• Efficiency: Remove all conversational fillers and greetings to save tokens.

Skill: Plan Tests

Purpose

Produces a Double-Loop TDD test plan that drives implementation order. With 95% line / 90% branch coverage enforced, the plan serves as the implementation roadmap — each test scenario maps to one Red-Green-Refactor cycle, ordered by Transformation Priority Premise (TPP).

Triggers

• /x-plan-tests STORY-ID — generate test plan for a specific story

Parameters

Parameter	Required	Description
STORY-ID	Yes	Story identifier to plan tests for. If not provided, prompt for it.

Workflow

0. PRE-CHECK       -> Verify existing plan freshness (mtime comparison)
1. GATHER CONTEXT  -> Subagent (model: opus) reads testing + architecture KPs, story, template, existing code
2. PLAN TESTS      -> Generate Double-Loop + TPP-ordered scenarios (inline)
3. ESTIMATE        -> Estimate coverage and validate completeness (inline)

Step 0 — Pre-Check: Idempotency (RULE-002 — Idempotency via Staleness Check)

Before generating a test plan, verify whether a valid plan already exists:

1. Resolve paths: Extract epic ID (XXXX) and story sequence (YYYY) from the story ID. Compute:

   • Story path: the story file provided as input
   • Plan path: ai/epics/epic-XXXX/plans/tests-story-XXXX-YYYY.md

2. Check existence: If the plan file does NOT exist, proceed to generation (Step 1).

3. Compare modification times: If the plan file exists:

   • If mtime(story file) <= mtime(plan file) — the plan is fresh. Log: "Reusing existing test plan from {date}" (where {date} is the plan file's last modified date). Return the existing plan. Do NOT invoke any subagent.
   • If mtime(story file) > mtime(plan file) — the plan is stale. Log: "Regenerating stale test plan for {story-id}". Proceed to generation (Step 1).

4. First generation: If the plan file does not exist at all, log: "Generating test plan for {story-id}". Proceed to generation (Step 1).

Step 1 — Gather Context (Subagent via Task, model: opus)

<!-- TELEMETRY: phase.start --> Bash command: $CLAUDE_PROJECT_DIR/.claude/hooks/telemetry-phase.sh start x-plan-tests Phase-1-KP-Read

Launch a single general-purpose subagent with explicit model: "opus" (Rule 23 RULE-002 — deep test planning quality for Double-Loop TDD + TPP):

Agent(
  subagent_type: "general-purpose",
  model: "opus",
  description: "Test Planning Assistant gathers context for test plan generation",
  prompt: "<see the blockquote below>"
)

You are a Test Planning Assistant gathering context for test plan generation.

Read the template for required output format (RULE-007):

• Read template at .claude/templates/_TEMPLATE-TEST-PLAN.md for required output format.
• If the template file does NOT exist, log: "Template not found, using inline format" and continue without it (RULE-012 — graceful fallback for projects without templates).

Read these knowledge packs:

• knowledge/testing/testing-philosophy.md — 8 test categories, fixture patterns, data uniqueness, async handling, real vs in-memory DB decisions
• knowledge/testing/testing-conventions.md — {{LANGUAGE}}-specific test frameworks, naming conventions, directory structure, assertion libraries
• knowledge/architecture/architecture-principles.md — exception hierarchy, layer boundaries (unit vs integration), dependency direction

Read testing knowledge pack for {{LANGUAGE}}-specific patterns:

• Read knowledge/testing.md for {{LANGUAGE}}-specific test frameworks, conventions, and patterns

Read the story: {STORY_PATH} Extract: acceptance criteria, sub-tasks, business rules, dependencies.

Scan existing code:

• List existing classes in target packages
• List existing test classes and patterns established

Return a structured context summary:

1. Test categories applicable to this story (from the 8 categories in testing-philosophy)
2. Naming convention: [method]_[scenario]_[expected] pattern with {{LANGUAGE}}-specific format
3. Test frameworks and assertion libraries to use (from testing-conventions)
4. Fixture pattern: final class + private constructor + static methods, naming a{Entity}()
5. Acceptance criteria extracted from story (each becomes ≥1 test)
6. Business rules extracted (each maps to parametrized tests)
7. Exception types in scope (each needs error path test)
8. Layer boundaries — which classes need unit vs integration tests
9. Existing patterns found in current test codebase
10. Contract tests applicable (if testing.contract_tests == true)
11. Chaos tests applicable (if testing.chaos_tests == true)
12. Template sections — if template was found, list the 8 mandatory sections to populate

<!-- TELEMETRY: phase.end --> Bash command: $CLAUDE_PROJECT_DIR/.claude/hooks/telemetry-phase.sh end x-plan-tests Phase-1-KP-Read ok

Step 2 — Generate Test Scenarios (Inline)

Using the context returned by the subagent, generate a Double-Loop TDD test plan. Organize scenarios by implementation order (TPP), NOT by test category.

<!-- TELEMETRY: phase.start --> Bash command: $CLAUDE_PROJECT_DIR/.claude/hooks/telemetry-phase.sh start x-plan-tests Phase-2-Acceptance-Tests

2.1 — Acceptance Tests (Outer Loop)

For each Gherkin scenario in the story, generate an acceptance test entry:

Field	Description
ID	AT-N (sequential)
Gherkin	Reference to the original scenario
Status	RED until all unit tests for this acceptance criteria complete
Components	List of classes/modules under test
Test Type	Integration, API, or E2E (depending on story scope)
Depends on	UT/IT IDs that must pass for this AT to go GREEN, or --
Parallel	yes if independent of other ATs

<!-- TELEMETRY: phase.end --> Bash command: $CLAUDE_PROJECT_DIR/.claude/hooks/telemetry-phase.sh end x-plan-tests Phase-2-Acceptance-Tests ok

2.2 — Unit Tests (Inner Loop, TPP Order)

<!-- TELEMETRY: phase.start --> Bash command: $CLAUDE_PROJECT_DIR/.claude/hooks/telemetry-phase.sh start x-plan-tests Phase-3-Unit-Tests-TPP

Generate unit test scenarios in strict TPP order. Each scenario represents one Red-Green-Refactor cycle.

TPP Level 1 — Degenerate Cases

• Null, empty, zero inputs — return default/error
• Transform: {}→nil or nil→constant

TPP Level 2 — Unconditional Paths

• Single valid input — direct output (no branching)
• Transform: constant→variable

TPP Level 3 — Simple Conditions

• Single if/else branching
• Transform: unconditional→conditional

TPP Level 4 — Complex Conditions

• Multiple branches, switch/match, compound boolean
• Transform: deeper conditional logic

TPP Level 5 — Iterations

• Collection processing, loops, map/filter/reduce
• Transform: scalar→collection, statement→recursion/iteration

TPP Level 6 — Edge Cases

• Boundary values (at-min, at-max, past-max)
• Transform: value→mutated value

For each unit test entry, include:

Field	Required	Description
ID	M	UT-N (sequential within TPP order)
Test	M	What to test (method + scenario + expected)
Implementation	O	Minimum code to pass this test
Transform	M	TPP transformation applied
TPP Level	M	1-6
Components	M	Classes/modules needed
Depends on	M	Previous test IDs that are prerequisites, or --
Parallel	M	yes if independent of other tests at same level

2.3 — Integration Tests (Cross-Component)

Position AFTER the unit tests of the components involved. Include only when multiple components interact (DB, HTTP, messaging).

Field	Required	Description
ID	M	IT-N (sequential)
Test	M	What to test
Components	M	Interacting components
Depends on	M	UT IDs that must pass first
Parallel	M	yes/no

2.4 — CRUD-Only Story Optimization

When a story describes a purely CRUD operation without branching logic:

• UTs should cover: degenerate (Level 1) — constant (Level 1) — variable (Level 2)
• Do NOT generate conditional or iteration UTs unless the business rules demand them
• Acceptance tests should focus on the full CRUD flow

<!-- TELEMETRY: phase.end --> Bash command: $CLAUDE_PROJECT_DIR/.claude/hooks/telemetry-phase.sh end x-plan-tests Phase-3-Unit-Tests-TPP ok

Step 3 — Estimate and Validate (Inline)

<!-- TELEMETRY: phase.start --> Bash command: $CLAUDE_PROJECT_DIR/.claude/hooks/telemetry-phase.sh start x-plan-tests Phase-4-Report

3.1 — Coverage Estimation Table

Class	Public Methods	Branches	Est. Tests	Line %	Branch %
[Name]	[count]	[count]	[count]	[%]	[%]

Flag any class where estimated coverage < 95% line / 90% branch.

3.2 — Quality Checks

1. Every Gherkin scenario maps to ≥1 acceptance test (AT)
2. Every acceptance criterion maps to ≥1 unit test chain (UT)
3. UT-1 is ALWAYS a degenerate case (TPP Level 1)
4. UTs follow non-decreasing TPP level order
5. Every exception has ≥1 error path test
6. Boundary values use triplet pattern (at-min, at-max, past-max)
7. Dependency markers are complete (no orphan UTs)
8. Estimated coverage meets thresholds (≥ 95% line, ≥ 90% branch)
9. Test naming follows convention: [method]_[scenario]_[expected]
10. No unnecessary UTs for CRUD-only stories (max Level 3 unless justified)

<!-- TELEMETRY: phase.end --> Bash command: $CLAUDE_PROJECT_DIR/.claude/hooks/telemetry-phase.sh end x-plan-tests Phase-4-Report ok

Output

Save to: ai/epics/epic-XXXX/plans/tests-story-XXXX-YYYY.md (extract epic ID XXXX and story sequence YYYY from the story ID). Ensure directory exists: mkdir -p ai/epics/epic-XXXX/plans.

Origin Marker (EPIC-0059 — mandatory)

Before writing the test plan file, prepend the YAML frontmatter block at the very top:

---
generated-by: x-plan-tests@$(git rev-parse HEAD 2>/dev/null || echo "unknown")
generated-at: $(date -u +%Y-%m-%dT%H:%M:%SZ)
story-id: ${STORY_ID}
---

This frontmatter is required by audit-execution-integrity.sh Phase-1 validation (EPIC-0059, Rule 24). Artifacts without this block fail the CI audit with EIE_EVIDENCE_MISSING.

# Test Plan — STORY-ID: [Title]

## Summary
- Acceptance tests: X (from Y Gherkin scenarios)
- Unit tests: ~Z (in TPP order)
- Integration tests: ~W
- Estimated line coverage: ~P%
- Estimated branch coverage: ~Q%

## Acceptance Tests (Outer Loop)

### AT-1: [Gherkin scenario name]
- **Gherkin**: [reference to story scenario]
- **Status**: RED until all unit tests complete
- **Components**: [list of components under test]
- **Acceptance Criteria**: [what must be true for this AT to pass]
- **Depends on**: UT-1, UT-2, ... (UTs that must pass)
- **Parallel**: yes/no

## Unit Tests (Inner Loop — TPP Order)

### UT-1: [degenerate case description] — TPP Level 1
- **Test**: [methodUnderTest]_[scenario]_[expectedBehavior]
- **Implementation**: [minimum code to pass]
- **Transform**: {}→nil
- **Components**: [class/module]
- **Depends on**: --
- **Parallel**: yes

### UT-2: [next case description] — TPP Level 2
- **Test**: [methodUnderTest]_[scenario]_[expectedBehavior]
- **Implementation**: [minimum code to pass]
- **Transform**: constant→variable
- **Components**: [class/module]
- **Depends on**: UT-1
- **Parallel**: no

...

## Integration Tests (Cross-Component)

### IT-1: [integration scenario]
- **Test**: [description]
- **Components**: [interacting components]
- **Depends on**: UT-3, UT-7
- **Parallel**: no

## Coverage Estimation
[table]

## Risks and Gaps
- [Hard-to-test scenarios]
- [Coverage gaps needing attention]

Anti-Patterns

• Do NOT write test code — only plan scenarios
• Do NOT organize test plan by category (Happy Path, Error Path, etc.) — use TPP ordering
• Do NOT skip error paths
• Do NOT forget boundary values (0, -1, max, empty, null)
• Do NOT plan tests for trivial getters/setters
• Do NOT ignore existing test patterns
• Do NOT create redundant tests covering the same branch

Error Handling

Scenario	Action
No story ID provided	Prompt user for story identifier
Story file not found	Abort with file-not-found message
Existing plan is fresh (RULE-002)	Return existing plan, skip generation
Template not found (RULE-012)	Log warning, use inline format, continue
No Gherkin scenarios in story	Generate tests from acceptance criteria text
No acceptance criteria found	Abort with warning, request story refinement

Template Fallback (RULE-012)

When .claude/templates/_TEMPLATE-TEST-PLAN.md is not available (projects predating EPIC-0024):

1. Log warning: "Template not found, using inline format"
2. Generate the test plan using the inline output format defined in the Output section above
3. Execution continues normally — no interruption, no error
4. The inline format produces the same 8 conceptual sections but without the template's strict structure

This ensures backward compatibility with projects that have not yet adopted template-based generation.

Knowledge Pack References

Pack	Files	Purpose
testing	knowledge/testing/testing-philosophy.md	8 test categories, fixture patterns, data uniqueness
testing	knowledge/testing/testing-conventions.md	{{LANGUAGE}}-specific test frameworks, naming, assertions
testing	knowledge/testing.md	{{LANGUAGE}}-specific test patterns
architecture	knowledge/architecture/architecture-principles.md	Exception hierarchy, layer boundaries

Integration Notes

Skill	Relationship	Context
x-implement-story	called-by	Invoked during Phase 1B
x-implement-task	reads	Output consumed as TDD roadmap

• Pre-check (RULE-002) prevents redundant regeneration when story has not changed
• Template reference (RULE-007) ensures consistent 8-section output format when available
• Subagent uses explicit model: "opus" (Rule 23 RULE-002) for deep test planning quality
• Output uses Double-Loop TDD format with TPP-ordered scenarios
• Output consumed by Phase 2 (developers) and Phase 3 (QA engineer validates coverage)
• Can be used standalone before any implementation task

Planning Status Propagation (Rule 22 / EPIC-0046)

V2-gated: only runs when SchemaVersionResolver.resolve(ai/epics/epic-XXXX/execution-state.json) == V2. v1 epics: skip silently (Rule 19).

After writing tests-story-XXXX-YYYY.md, check the associated story's lifecycle status. The test plan, like the architecture plan, is a secondary writer — the primary transition Pendente → Planejada is owned by x-plan-story. When x-plan-tests runs standalone (without x-plan-story), it promotes the story from Pendente to Planejada so the source artifact reflects that the test plan is in place.

Steps (end of test-plan generation, BEFORE the final commit):

1. Detect v2 via SchemaVersionResolver. If v1: skip.

2. Read current story status:

   CURRENT=$(java -cp $CLAUDE_PROJECT_DIR/java/target/classes \
       dev.iadev.adapter.inbound.cli.StatusFieldParserCli \
       read ai/epics/epic-XXXX/story-XXXX-YYYY.md)
   

3. If CURRENT == "Pendente" → write Planejada. Idempotent when already Planejada.

4. Stage and commit:

   git add ai/epics/epic-XXXX/story-XXXX-YYYY.md ai/epics/epic-XXXX/plans/tests-story-XXXX-YYYY.md
   

   Skill(skill: "x-commit-changes", args: "docs(story-XXXX-YYYY): add test plan + update status to Planejada")
   

Fail-loud: non-zero CLI exit aborts the skill (RULE-046-08).