Ln 520 Test Planner

> Paths: File paths (references/, ../ln-*) are relative to this skill directory.

Test Planning Orchestrator

Type: L2 Coordinator Category: 5XX Quality

Runtime-backed test-planning coordinator. The runtime owns skip/reuse gates, worker summary tracking, and deterministic resume.

Inputs

| Input | Required | Source | Description | |-------|----------|--------|-------------| | storyId | Yes | args, git branch, kanban, user | Story to process | | --simplified | No | args | Skip research (ln-521) and manual testing (ln-522). Run only auto-test planning (ln-523). Used in fast-track mode. |

Resolution: Story Resolution Chain. Status filter: To Review

Purpose & Scope

• Orchestrate test planning: research → manual testing → automated test planning
• Delegate to workers: ln-521-test-researcher, ln-522-manual-tester, ln-523-auto-test-planner
• No direct work — only coordination and delegation via Skill tool

Runtime Contract

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

Runtime family: test-planning-runtime

Identifier:

• Story ID

Phases:

1. PHASE_0_CONFIG
2. PHASE_1_DISCOVERY
3. PHASE_2_RESEARCH
4. PHASE_3_MANUAL_TESTING
5. PHASE_4_AUTO_TEST_PLANNING
6. PHASE_5_FINALIZE
7. PHASE_6_SELF_CHECK

Worker summary contract:

• ln-521, ln-522, ln-523 receive deterministic child runId plus exact summaryArtifactPath
• checkpoint child runtime metadata before waiting for the artifact
• each worker writes a test-planning-worker summary envelope before terminal outcome
• ln-520 consumes worker summaries, not free-text worker prose

When to Use

This skill should be used when:

• Story passed implementation and regression work and needs full test planning
• All implementation tasks in Story are Done
• Need complete test planning (research + manual + auto)

Prerequisites:

• All implementation Tasks in Story status = Done
• Regression tests passed (ln-513)
• Code quality checked (ln-511)

Pipeline Overview

ln-520-test-planner (Orchestrator)
    │
    ├─→ ln-521-test-researcher
    │     └─→ Posts "## Test Research: {Feature}" comment
    │
    ├─→ ln-522-manual-tester
    │     └─→ Creates tests/manual/ scripts + "## Manual Testing Results" comment
    │
    └─→ ln-523-auto-test-planner
          └─→ Creates test task via the configured tracker provider via ln-301/ln-302

Workflow

Phase 0: Resolve Inputs

MANDATORY READ: Load references/input_resolution_pattern.md

1. Resolve storyId: Run Story Resolution Chain per guide (status filter: [To Review]).

Phase 1: Discovery

1) Auto-discover Team ID from docs/tasks/kanban_board.md 2) Validate Story ID

Phase 2: Research Delegation

> Simplified mode gate: > - IF --simplified flag AND research comment already exists on Story: Skip Phase 2 (research). Proceed to Phase 4. > - IF --simplified flag AND no research comment: Skip Phase 2. Proceed to Phase 4 (ln-523 will generate minimal inline research).

1) Check if research exists:

• Search tracker comments for "## Test Research:" header (listComments)
• If found → skip to Phase 3

2) If no research:

• Compute deterministic child run inputs for ln-521
• Start test-planning-worker-runtime and checkpoint child_run before delegation
• Invoke ln-521-test-researcher via managed worker run
• Pass: Story ID
• Wait for completion
• Read only the resulting test-planning-worker artifact
• Verify research comment created

Phase 3: Manual Testing Delegation

> Simplified mode gate: > - IF --simplified flag: Skip Phase 3 (manual testing). Proceed to Phase 4.

1) Check if manual testing done:

• Search tracker comments for "## Manual Testing Results" header (listComments)
• If found with all AC passed → skip to Phase 4

2) If manual testing needed:

• Compute deterministic child run inputs for ln-522
• Start test-planning-worker-runtime and checkpoint child_run before delegation
• Invoke ln-522-manual-tester via managed worker run
• Pass: Story ID
• Wait for completion
• Read only the resulting test-planning-worker artifact
• Verify results comment created

3) If any AC failed:

• Stop pipeline
• Report to ln-500: "Manual testing failed, Story needs fixes"

Phase 4: Auto Test Planning Delegation

1) Invoke auto test planner:

• Compute deterministic child run inputs for ln-523
• Start test-planning-worker-runtime and checkpoint child_run before delegation
• Invoke ln-523-auto-test-planner via managed worker run
• Pass: Story ID
• Wait for completion
• Read only the resulting test-planning-worker artifact

2) Verify results:

• Test task created via the configured tracker provider (or updated if existed)
• Return task URL to ln-500

Phase 5: Report to Caller

1) Return summary to ln-500:

• Research: completed / skipped (existed)
• Manual testing: passed / failed
• Test task: created / updated + URL

Runtime Summary Artifact

MANDATORY READ: Load references/coordinator_summary_contract.md

Write .hex-skills/runtime-artifacts/runs/{run_id}/story-tests/{story_id}.json before finishing.

Worker Invocation (MANDATORY)

Host Skill Invocation: Skill(skill: "...", args: "...") is mandatory delegation.

• Claude: call the Skill tool exactly as shown.
• Codex: if no Skill tool exists, locate the named skill in available skills, read its SKILL.md, treat args as $ARGUMENTS, execute that skill workflow, then return here with its result/artifact.
• Do not inline worker logic or mark the worker complete without executing the target skill.

> CRITICAL: All delegations use Agent tool with subagent_type: "general-purpose" for context isolation, but every managed worker run must be prepared through test-planning-worker-runtime first.

| Phase | Worker | Purpose | |-------|--------|---------| | 2 | ln-521-test-researcher | Runtime-backed managed Agent call; artifact is the only completion signal | | 3 | ln-522-manual-tester | Runtime-backed managed Agent call; artifact is the only completion signal | | 4 | ln-523-auto-test-planner | Runtime-backed managed Agent call; artifact is the only completion signal |

Prompt template:

Agent(description: "[Phase N] test planning via ln-52X",
     prompt: "Execute test planning worker.

Step 1: Start worker runtime:
  node references/scripts/test-planning-worker-runtime/cli.mjs start --skill {worker} --story {storyId} --manifest-file .hex-skills/test-planning/{worker}--{storyId}_manifest.json --run-id {childRunId} --summary-artifact-path .hex-skills/runtime-artifacts/runs/{parent_run_id}/test-planning-worker/{worker}--{storyId}.json

Step 2: Checkpoint child metadata:
  node references/scripts/test-planning-runtime/cli.mjs checkpoint --story {storyId} --phase PHASE_N --payload '{\"child_run\":{\"worker\":\"{worker}\",\"run_id\":\"{childRunId}\",\"summary_artifact_path\":\".hex-skills/runtime-artifacts/runs/{parent_run_id}/test-planning-worker/{worker}--{storyId}.json\"}}'

Step 3: Invoke worker:
  Skill(skill: \"ln-52X-{worker}\", args: \"{storyId} --run-id {childRunId} --summary-artifact-path .hex-skills/runtime-artifacts/runs/{parent_run_id}/test-planning-worker/{worker}--{storyId}.json\")

CONTEXT:
Story: {storyId}",
     subagent_type: "general-purpose")

Anti-Patterns:

• ❌ Direct Agent invocation without child runtime start + child_run checkpoint
• ❌ Running web searches directly (delegate to ln-521)
• ❌ Creating bash test scripts directly (delegate to ln-522)
• ❌ Creating test tasks directly (delegate to ln-523)
• ❌ Skipping any phase without justification

TodoWrite format (mandatory)

- Resolve Story and prerequisites (pending)
- Check or reuse research state (pending)
- Start ln-521 child runtime, checkpoint metadata, or skip deterministically (pending)
- Check or reuse manual testing state (pending)
- Start ln-522 child runtime, checkpoint metadata, or skip deterministically (pending)
- Start ln-523 child runtime, checkpoint metadata, and verify test-task result (pending)
- Write story-tests summary artifact (pending)
- Report final planning outcome (pending)

Critical Rules

• No direct work: Orchestrator only delegates, never executes tasks itself
• Sequential execution: 521 → 522 → 523 (each depends on previous)
• Fail-fast: If manual testing fails, stop pipeline and report
• Skip detection: Check for existing comments before invoking workers
• Single responsibility: Each worker does one thing well
• Every managed worker run must start through test-planning-worker-runtime before the Agent wrapper invokes the worker.

Definition of Done

• [ ] Story ID validated
• [ ] Research phase: ln-521 invoked OR existing comment found
• [ ] Manual testing phase: ln-522 invoked OR existing results found
• [ ] Auto test planning phase: ln-523 invoked
• [ ] Child test-planning runtimes started and checkpointed before every managed worker invoke
• [ ] Test task created/updated via the configured tracker provider
• [ ] Summary prepared with phase results and test task URL
• [ ] Story-test summary artifact written to the shared location

Output: Summary with phase results + test task URL

Phase 6: Meta-Analysis

Optional reference: load references/meta_analysis_protocol.md only when the user asks for post-run meta-analysis or protocol-formatted run reflection.

Skill type: planning-coordinator. When requested, run after all phases complete. Output to chat using the planning-coordinator format.

Reference Files

• Workers: ../ln-521-test-researcher/SKILL.md, ../ln-522-manual-tester/SKILL.md, ../ln-523-auto-test-planner/SKILL.md
• Risk-based testing: references/risk_based_testing_guide.md

---

Version: 4.0.0 Last Updated: 2026-01-15

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: levnikolaevich
• Source: levnikolaevich/claude-code-skills
• License: MIT
• Homepage: https://levnikolaevich.github.io/claude-code-skills/

Install and usage instructions live in the source repository linked above.