ln-500-story-quality-gate（ln 500 层质量门）

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}.

Type: L2 Coordinator Category: 5XX Quality

Story Quality Gate

Runtime-backed gate coordinator. Owns fast-track routing, quality/test summaries, final Story verdict, and branch finalization.

Inputs

Resolution: Story Resolution Chain. Status filter: To Review

Purpose & Scope

• Invoke ln-510-quality-coordinator
• Invoke ln-520-test-planner when needed
• Wait deterministically for test-task readiness
• Calculate gate verdict: PASS | CONCERNS | FAIL | WAIVED
• Move Story to Done only on passing outcomes
• Persist resumable gate runtime in .hex-skills/story-gate/runtime/

Runtime Contract

MANDATORY READ: Load shared/references/environment_state_contract.md, shared/references/storage_mode_detection.md, shared/references/input_resolution_pattern.md MANDATORY READ: Load shared/references/coordinator_runtime_contract.md, shared/references/story_gate_runtime_contract.md, shared/references/coordinator_summary_contract.md, shared/references/loop_health_contract.md MANDATORY READ: Load shared/references/git_worktree_fallback.md MANDATORY READ: Load references/minimum_quality_checks.md

Runtime CLI:

node shared/scripts/story-gate-runtime/cli.mjs start --story {storyId} --manifest-file .hex-skills/story-gate/manifest.json
node shared/scripts/story-gate-runtime/cli.mjs status
node shared/scripts/story-gate-runtime/cli.mjs record-quality --payload '{...}'
node shared/scripts/story-gate-runtime/cli.mjs record-test-status --payload '{...}'
node shared/scripts/story-gate-runtime/cli.mjs record-stage-summary --story {storyId} --payload '{...}'
node shared/scripts/story-gate-runtime/cli.mjs checkpoint --phase PHASE_6_VERDICT --payload '{...}'
node shared/scripts/story-gate-runtime/cli.mjs advance --to PHASE_7_FINALIZATION

4-Level Gate Model

Workflow

Phase 0: Config

1. Resolve storyId and task_provider.
2. Check for prior Stage 3 artifact at .hex-skills/runtime-artifacts/runs/*/story-quality/{storyId}.json:

- If found with verdict=FAIL, load payload.metadata.rework_hint as previous_cycle context. - Record cycle_number (1-based, incremented from prior artifact count).

1. Build gate manifest:

- story_id - task_provider - project_root - worktree_dir - branch - fast_track_policy - nfr_policy - test_task_policy - previous_cycle (null or loaded rework_hint) - cycle_number

1. Start runtime and checkpoint PHASE_0_CONFIG.

Phase 1: Discovery

1. Load Story metadata and child task metadata.
2. Detect existing test task and its current status.
3. Capture readiness inputs if available from upstream pipeline.
4. Checkpoint PHASE_1_DISCOVERY.

Phase 2: Fast-Track

1. Determine fast_track=true only when readiness explicitly allows it.
2. Checkpoint PHASE_2_FAST_TRACK with:

- fast_track - gate scope summary

Phase 3: Quality Checks

1. Compute:

- childRunId = {parent_run_id}--ln-510--{storyId} - childSummaryArtifactPath =.hex-skills/runtime-artifacts/runs/{parent_run_id}/story-quality/{storyId}.json

1. Materialize child manifest and start child coordinator runtime:

- node shared/scripts/quality-runtime/cli.mjs start --story {storyId} --manifest-file.hex-skills/story-gate/ln-510--{storyId}_manifest.json --run-id {childRunId}

1. Checkpoint PHASE_3_QUALITY_CHECKS with:

- child_run.worker=ln-510 - child_run.run_id={childRunId} - child_run.summary_artifact_path={childSummaryArtifactPath} - child_run.phase_context=quality_checks

1. Invoke ln-510-quality-coordinator with managed transport inputs:

- full mode: Skill(skill: "ln-510-quality-coordinator", args: "{storyId} --run-id {childRunId} --summary-artifact-path {childSummaryArtifactPath}") - fast-track: Skill(skill: "ln-510-quality-coordinator", args: "{storyId} --fast-track --run-id {childRunId} --summary-artifact-path {childSummaryArtifactPath}") - full mode with prior cycle: append --previous-cycle-focus "{blocking_categories}" when previous_cycle is not null

1. Read child story-quality artifact only, then record-quality.
2. Before another quality/rework cycle, compare FAIL evidence against prior cycle:

- progress = new quality artifact, new fix tasks, code delta, status delta, or changed blocking category - same FAIL with no new evidence = record loop health and pause before another rework cycle

1. Checkpoint PHASE_3_QUALITY_CHECKS with the recorded quality summary.
2. If the quality summary already implies hard FAIL, you may jump directly to PHASE_6_VERDICT.

Phase 4: Test Planning

1. Decide whether planning is needed:

- no test task -> invoke ln-520 - fast-track -> invoke simplified ln-520 - test task already exists and is terminal (Done | SKIPPED | VERIFIED) -> checkpoint as reused

1. When invoking ln-520, compute:

- childRunId = {parent_run_id}--ln-520--{storyId} - childSummaryArtifactPath =.hex-skills/runtime-artifacts/runs/{parent_run_id}/story-tests/{storyId}.json

1. Materialize child manifest and start child coordinator runtime:

- node shared/scripts/test-planning-runtime/cli.mjs start --story {storyId} --manifest-file.hex-skills/story-gate/ln-520--{storyId}_manifest.json --run-id {childRunId}

1. Checkpoint PHASE_4_TEST_PLANNING with:

- child_run.worker=ln-520 - child_run.run_id={childRunId} - child_run.summary_artifact_path={childSummaryArtifactPath} - child_run.phase_context=test_planning

1. Invoke ln-520-test-planner with managed transport inputs:

- normal mode: Skill(skill: "ln-520-test-planner", args: "{storyId} --run-id {childRunId} --summary-artifact-path {childSummaryArtifactPath}") - simplified mode: Skill(skill: "ln-520-test-planner", args: "{storyId} --simplified --run-id {childRunId} --summary-artifact-path {childSummaryArtifactPath}")

1. Read child story-tests artifact only, then record-test-status.
2. Checkpoint PHASE_4_TEST_PLANNING.

Phase 5: Test Verification

1. If test task exists but is not Done, pause runtime:

- phase = PAUSED - resume_action = wait for test task completion

1. When resumed, verify:

- test task terminal status is Done, SKIPPED, or VERIFIED - coverage summary exists - planned scenarios and Story AC coverage are machine-readable

1. Checkpoint PHASE_5_TEST_VERIFICATION with:

- test_task_status - verification result

Phase 6: Verdict

1. Calculate quality_score.
2. Evaluate NFR validation:

- full gate: security, performance, reliability, maintainability - fast-track: security mandatory, others may downgrade to concerns-only scope

1. End-to-End Scenario Completeness Walkthrough: For each AC where an actor (user, bot, scheduler, handler, pipeline) must invoke or consume a mechanism, trace all 5 segments of the interaction path:

- (1) Actor trigger — what initiates the scenario (user sends message, timer fires, webhook arrives, event is enqueued) - (2) Entry point — the named mechanism (MCP tool, API endpoint, CLI command, UI component, chat handler, config file, cron handler) - (3) Discovery — how the actor's system finds/loads the mechanism at runtime (config registration, route mounting, plugin loading, system prompt, environment variable) - (4) Usage context — what the actor's system needs to correctly invoke the mechanism (instructions, prompts, schemas, type hints, documentation, parameter guidance) - (5) Observable outcome — the verifiable result (response message, state change, log entry, notification) If any segment is missing, create an AC- prefixed issue. Common failures: infrastructure exists but nothing exposes it (missing segment 2), mechanism exists but the actor's system can't find it (missing segment 3), mechanism is discoverable but the actor's system doesn't know when or how to use it (missing segment 4). If any AC- issue is found: verdict MUST be FAIL; create fix tasks for the missing segments.

1. Determine final verdict.
2. For FAIL:

- create follow-up tasks - keep Story out of Done

1. Checkpoint PHASE_6_VERDICT with:

- final_result - quality_score - nfr_validation - fix_tasks_created

Phase 7: Finalization

For PASS | CONCERNS | WAIVED:

1. Commit and push verified branch if needed.
2. Move Story to Done.
3. Post gate comment.
4. Cleanup worktree when caller does not own it.

For FAIL:

1. Do not finalize branch as accepted.
2. Create follow-up tasks from blocking issues.
3. Checkpoint PHASE_7_FINALIZATION with status=requires_rework.
4. Record resulting Story status, follow-up task IDs.

After finalization, write a Stage 3 coordinator artifact with:

• summary_kind=pipeline-stage
• stage=3
• story_id
• status=completed
• final_result
• story_status
• verdict
• quality_score
• warnings
• metadata.rework_hint — when verdict is FAIL, include:

- rework_tasks: list of created follow-up task IDs - blocking_categories: list of issue categories that caused FAIL (e.g. ac_gap, security, regression) - suggested_focus: concise one-line description of what the rework cycle should prioritize

Phase 8: Self-Check

Build final checklist from runtime state:

• Config, discovery, and fast-track checkpoints exist
• Quality summary recorded from ln-510
• Test-planning and test-verification state are deterministic
• Final verdict checkpoint exists
• Story final status recorded
• Branch finalization recorded or skipped by verdict
• Stage 3 coordinator artifact recorded

Checkpoint PHASE_8_SELF_CHECK with pass=true|false. Complete runtime only after pass=true.

Worker Invocation (MANDATORY)

childRunId = "{parent_run_id}--ln-510--{storyId}"
childSummaryArtifactPath = ".hex-skills/runtime-artifacts/runs/{parent_run_id}/story-quality/{storyId}.json"
node shared/scripts/quality-runtime/cli.mjs start --story {storyId} --manifest-file .hex-skills/story-gate/ln-510--{storyId}_manifest.json --run-id {childRunId}
node shared/scripts/story-gate-runtime/cli.mjs checkpoint --phase PHASE_3_QUALITY_CHECKS --payload '{"child_run":{"worker":"ln-510","run_id":"{childRunId}","summary_artifact_path":"{childSummaryArtifactPath}","phase_context":"quality_checks"}}'
Skill(skill: "ln-510-quality-coordinator", args: "{storyId} --run-id {childRunId} --summary-artifact-path {childSummaryArtifactPath}")

childRunId = "{parent_run_id}--ln-520--{storyId}"
childSummaryArtifactPath = ".hex-skills/runtime-artifacts/runs/{parent_run_id}/story-tests/{storyId}.json"
node shared/scripts/test-planning-runtime/cli.mjs start --story {storyId} --manifest-file .hex-skills/story-gate/ln-520--{storyId}_manifest.json --run-id {childRunId}
node shared/scripts/story-gate-runtime/cli.mjs checkpoint --phase PHASE_4_TEST_PLANNING --payload '{"child_run":{"worker":"ln-520","run_id":"{childRunId}","summary_artifact_path":"{childSummaryArtifactPath}","phase_context":"test_planning"}}'
Skill(skill: "ln-520-test-planner", args: "{storyId} --run-id {childRunId} --summary-artifact-path {childSummaryArtifactPath}")

TodoWrite format (mandatory)

- Start ln-500 runtime (pending)
- Load Story/test-task metadata (pending)
- Decide fast-track mode (pending)
- Start ln-510 child runtime, checkpoint child_run, and record quality summary (pending)
- Start or reuse ln-520 child runtime, checkpoint child_run, and record test-planning summary (pending)
- Verify test task readiness (pending)
- Calculate final verdict (pending)
- Finalize Story/branch state and Stage 3 artifact (pending)
- Run runtime self-check and complete (pending)

Critical Rules

• Runtime state is the gate orchestration SSOT.
• ln-510 and ln-520 are consumed only through summary JSON artifacts.
• Child coordinator runtime status is resume-only metadata; phase completion still depends on the recorded child artifact.
• Test-task waiting is a deterministic pause, not an implicit stop.
• FAIL is a valid terminal gate result if follow-up actions are recorded correctly.
• ln-1000 consumes the Stage 3 coordinator artifact, not free-text stage output.
• Story may go to Done only on PASS, CONCERNS, or WAIVED.

Definition of Done

• Runtime started and config/discovery checkpoints recorded
• Fast-track decision checkpointed
• ln-510 summary recorded in runtime
• ln-520 summary recorded or reused deterministically
• Test verification reached terminal state or deterministic pause
• Final verdict checkpointed with quality score and NFR results
• Story moved to Done only for passing outcomes, or follow-up tasks created for FAIL
• Stage 3 coordinator artifact recorded before completion
• Self-check passed and runtime completed

Phase 9: Meta-Analysis

MANDATORY READ: Load shared/references/meta_analysis_protocol.md

Skill type: execution-orchestrator. Run after phases complete. Output to chat using the execution-orchestrator format.

Reference Files

• shared/references/coordinator_runtime_contract.md
• shared/references/story_gate_runtime_contract.md
• shared/references/coordinator_summary_contract.md
• references/minimum_quality_checks.md
• ../ln-510-quality-coordinator/SKILL.md
• ../ln-520-test-planner/SKILL.md

Version: 7.0.0 Last Updated: 2026-02-09