ln-523-auto-test-planner（ln 523 自动测试计划器）

Paths: File paths (shared/, references/, ../ln-*) are relative to skills repo root. If not found at CWD, locate this SKILL.md directory and go up one level for repo root. If shared/ is missing, fetch files via WebFetch from https://raw.githubusercontent.com/levnikolaevich/claude-code-skills/master/skills/{path}.

Automated Test Planner

Type: L3 Worker

Creates Story test task with comprehensive automated test coverage (E2E/Integration/Unit) based on Risk-Based Testing methodology and REAL manual testing results.

Inputs

Resolution: Story Resolution Chain. Status filter: To Review

Purpose & Scope

• Create comprehensive test task for Story automation
• Calculate risk-based priorities (Impact x Probability)
• Generate 11-section test plan from manual test results
• Delegate to ln-301-task-creator (CREATE) or ln-302-task-replanner (REPLAN)
• NOT for: manual testing, research, or orchestration

When to Use

• Use when implementation work is done and automated test coverage must be planned
• All implementation Tasks in Story status = Done
• ln-521 research: uses if available, generates minimal inline research if missing
• ln-522 manual testing: uses if available, marks as 'skipped by policy' if missing

Automation: Supports autoApprove: true in managed runs to skip manual confirmation.

Workflow

Phase 0: Tools Config

MANDATORY READ: Load shared/references/environment_state_contract.md, shared/references/storage_mode_detection.md, and shared/references/input_resolution_pattern.md

Extract: task_provider = Task Management → Provider (linear | file).

Phase 1: Discovery (Automated)

1. Resolve storyId: Run Story Resolution Chain per guide (status filter: [To Review]).
2. Auto-discover Team ID from docs/tasks/kanban_board.md (see CLAUDE.md "Configuration Auto-Discovery")

Phase 2: Story + Tasks Analysis (NO Dialog)

Step 0: Study Project Test Files

1. Scan for test-related files:

- tests/README.md (commands, setup, environment) - Test configs (jest.config.js, vitest.config.ts, pytest.ini) - Existing test structure (tests/, tests/ directories) - Coverage config (.coveragerc, coverage.json)

1. Extract: test commands, framework, patterns, coverage thresholds
2. Ensures test planning aligns with project practices

Step 1: Load Research and Manual Test Results

1. Fetch Story (must have label "user-story"):

- IF task_provider = linear: get_issue(storyId) — extract Story.id (UUID, NOT short ID) - IF task_provider = file: Read story.md — extract Story metadata

1. Load research comment (from ln-521): "## Test Research: {Feature}"

- IF task_provider = linear: list_comments(issueId=storyId) → find matching comment - IF task_provider = file: Glob("docs/tasks/epics/*/stories/*/comments/*.md") → find matching comment - IF research comment not found: generate 3-5 bullet points of key test areas based on Story AC and task descriptions (inline, no external research).

1. Load manual test results comment (from ln-522): "## Manual Testing Results"

- Same approach as research comment above - IF manual test results not found: skip manual test coverage analysis. Note in output: 'Manual testing: skipped by policy (simplified mode).'

1. Parse sections: AC results (PASS/FAIL), Edge Cases, Error Handling, Integration flows
2. Map to test design: PASSED AC -> E2E, Edge cases -> Unit, Errors -> Error handling, Flows -> Integration

Step 2: Analyze Story + Tasks

1. Parse Story: Goal, Test Strategy, Technical Notes
2. Fetch all child Tasks (status = Done):

- IF task_provider = linear: list_issues(parentId=Story.id, state="Done") - IF task_provider = file: Glob("docs/tasks/epics/*/stories/*/tasks/*.md") → filter by **Status:** Done

1. Analyze each Task:

- Components implemented - Business logic added - Integration points created - Conditional branches (if/else/switch)

1. Identify what needs testing

Phase 3: Parsing Strategy for Manual Test Results

Process: Locate Linear comment with "Manual Testing Results" header -> Verify Format Version 1.0 -> Extract structured sections (Acceptance Criteria, Test Results by AC, Edge Cases, Error Handling, Integration Testing) using regex -> Validate (at least 1 PASSED AC, AC count matches Story, completeness check) -> Map parsed data to test design structure

Error Handling: Missing comment -> use fallback (inline research or skip per Phase 2 logic), Missing format version -> WARNING (try legacy parsing), Required section missing -> use fallback, No PASSED AC -> ERROR (fix implementation)

Phase 4: Risk-Based Test Planning (Automated)

MANDATORY READ: Load shared/references/risk_based_testing_guide.md for complete methodology.

E2E-First Approach: Prioritize by business risk (Priority = Impact x Probability), not coverage metrics.

Workflow:

Step 1: Risk Assessment

Calculate Priority for each scenario from manual testing:

Priority = Business Impact (1-5) x Probability (1-5)

Decision Criteria:

• Priority ≥15 -> MUST test
• Priority 9-14 -> SHOULD test if not covered
• Priority <=8 -> SKIP (manual testing sufficient)

Step 2: E2E Test Selection (2-5): Baseline 2 (positive + negative) ALWAYS + 0-3 additional (Priority ≥15 only)

Step 3: Unit Test Selection (0-15): DEFAULT 0. Add ONLY for complex business logic (Priority ≥15): financial, security, algorithms

Step 4: Integration Test Selection: DEFAULT 0. Add ONLY if E2E gaps AND Priority ≥15: rollback, concurrency, external API errors

Step 5: Validation: Each test passes Usefulness Criteria (Priority ≥15, Confidence ROI, Behavioral, Predictive, Specific, Non-Duplicative)

Phase 5: Test Task Generation (Automated)

Generates complete test task per test_task_template.md (11 sections):

Sections 1-7: Context, Risk Matrix, E2E/Integration/Unit Tests (with Priority scores + justifications), Coverage, DoD

Section 8: Existing Tests to Fix (analysis of affected tests from implementation tasks)

Section 9: Infrastructure Changes (packages, Docker, configs - based on test dependencies)

Section 10: Documentation Updates (README, CHANGELOG, tests/README, config docs)

Section 11: Legacy Code Cleanup (unsupported patterns, backward compat, dead code)

Shows preview for review.

Phase 6: Confirmation & Delegation

Step 1: Preview generated test plan (always displayed for transparency)

Step 2: Confirmation logic:

• autoApprove: true -> proceed automatically
• Manual run -> prompt user to type "confirm"

Step 3: Check for existing test task

• IF task_provider = linear: list_issues(parentId=Story.id, label="tests")
• IF task_provider = file: Glob("docs/tasks/epics/*/stories/*/tasks/*.md") → filter by **Labels:** containing tests

Decision:

• Count = 0 -> CREATE MODE (Step 4a)
• Count >= 1 -> REPLAN MODE (Step 4b)

Step 4a: CREATE MODE (if Count = 0)

Managed delegation sequence:

1. Compute childRunId = {parent_run_id}--ln-301--{storyId}
2. Compute childSummaryArtifactPath =.hex-skills/runtime-artifacts/runs/{parent_run_id}/task-plan/ln-301--{storyId}.json
3. Materialize child manifest at .hex-skills/test-planning/ln-301--{storyId}_manifest.json
4. Start task-plan-worker-runtime with both transport inputs
5. Checkpoint child_run metadata in PHASE_6_DELEGATE_TASK_PLAN
6. Invoke ln-301-task-creator with --run-id and --summary-artifact-path
7. Read only the resulting child task-plan artifact

Pass to worker:

• taskType, teamId, storyData (Story.id, title, AC, Technical Notes, Context)
• researchFindings (from ln-521 comment)
• manualTestResults (from ln-522 comment)
• testPlan (e2eTests, integrationTests, unitTests, riskPriorityMatrix)
• infrastructureChanges, documentationUpdates, legacyCleanup

Worker returns: Task URL + task-plan artifact

Step 4b: REPLAN MODE (if Count >= 1)

Managed delegation sequence:

1. Compute childRunId = {parent_run_id}--ln-302--{storyId}
2. Compute childSummaryArtifactPath =.hex-skills/runtime-artifacts/runs/{parent_run_id}/task-plan/ln-302--{storyId}.json
3. Materialize child manifest at .hex-skills/test-planning/ln-302--{storyId}_manifest.json
4. Start task-plan-worker-runtime with both transport inputs
5. Checkpoint child_run metadata in PHASE_6_DELEGATE_TASK_PLAN
6. Invoke ln-302-task-replanner with --run-id and --summary-artifact-path
7. Read only the resulting child task-plan artifact

Pass to worker:

• Same data as CREATE MODE + existingTaskIds

Worker returns: Operations summary + child task-plan artifact

Step 5: Return structured summary to caller

Runtime Summary Artifact

MANDATORY READ: Load shared/references/test_planning_summary_contract.md, shared/references/test_planning_worker_runtime_contract.md, shared/references/task_plan_worker_runtime_contract.md

Runtime profile:

• family: test-planning-worker
• worker: ln-523
• summary kind: test-planning-worker
• payload fields used by coordinators: worker, status, warnings, test_task_id, test_task_url, coverage_summary, planned_scenarios

Invocation rules:

• standalone: omit runId and summaryArtifactPath
• managed: pass both runId and exact summaryArtifactPath
• always write the validated summary before terminal outcome

Delegated child worker rules:

• when delegating to ln-301 or ln-302, ln-523 becomes the parent runtime
• start task-plan-worker-runtime before delegation
• checkpoint child_run metadata before the child Skill call
• consume only the child task-plan artifact, never child prose

Definition of Done

Research and Manual Results Loaded:

• Research comment "## Test Research: {Feature}" found (from ln-521)
• Manual test results "## Manual Testing Results" found (from ln-522)
• At least 1 AC marked as PASSED

Risk-Based Test Plan Generated:

• Risk Priority Matrix calculated for all scenarios
• E2E tests: Baseline positive + negative, additional only if Priority ≥15
• Integration tests: ONLY if E2E doesn't cover AND Priority ≥15
• Unit tests: ONLY complex business logic with Priority ≥15
• Each test passes all 6 Usefulness Criteria
• No framework/library testing: Each test validates OUR business logic only

Test Task Description Complete (11 sections):

• All 11 sections populated per template
• Risk Priority Matrix included
• Each test beyond baseline 2 justified

Worker Delegation Executed:

• CREATE MODE: Delegated to ln-301-task-creator through task-plan-worker-runtime
• REPLAN MODE: Delegated to ln-302-task-replanner through task-plan-worker-runtime
• Child run metadata checkpointed before delegation
• Child task-plan artifact consumed before final summary
• Linear Issue URL returned

Output:

• CREATE MODE: Linear Issue URL + confirmation
• REPLAN MODE: Operations summary + URLs

Reference Files

• Environment state: shared/references/environment_state_contract.md
• Storage mode operations: shared/references/storage_mode_detection.md
• Risk-based testing methodology: shared/references/risk_based_testing_guide.md
• Auto-discovery patterns: shared/references/auto_discovery_pattern.md
• Test task template: shared/templates/test_task_template.md (workers ln-301/ln-302 load via Template Loading)
• Testing examples: references/risk_based_testing_examples.md
• MANDATORY READ: Load shared/references/research_tool_fallback.md

Critical Rules

• Manual results preferred: Use ln-522 manual testing results when available. When missing (simplified mode), generate minimal inline research and mark manual testing as skipped
• E2E-first, not unit-first: Baseline is always 2 E2E (positive + negative); unit/integration added only for Priority >= 15
• No framework testing: Every test must validate OUR business logic; never test library/framework behavior
• Usefulness enforcement: Every test beyond baseline must pass all 6 Usefulness Criteria (see risk_based_testing_guide.md)
• Delegate, don't create: Task creation goes through ln-301/ln-302 workers; this skill generates the plan only
• Managed child runs only: Parent-child delegation to ln-301/ln-302 must use task-plan-worker-runtime, deterministic child runId, exact child summaryArtifactPath, and artifact-only consumption

Best Practices

Minimum Viable Testing: Start with baseline E2E (positive + negative). Each additional test must pass all 6 Usefulness Criteria.

Risk-Based Testing: Prioritize by Business Impact x Probability. E2E-first from ACTUAL manual testing results. Priority ≥15 scenarios covered by tests.

Expected-Based Testing: For deterministic tests, compare actual vs expected using diff. MANDATORY READ: Load ../ln-522-manual-tester/SKILL.md — section "Test Design Principles".

Version: 1.0.0 Last Updated: 2026-01-15