Story Quality Gate

Two-pass Story review with 4-level Gate verdict, Quality Score calculation, and NFR validation (security, performance, reliability, maintainability).

Purpose & Scope

• •Pass 1 (after impl tasks Done): run code-quality, lint, regression, and manual testing; if all pass, create/confirm test task; otherwise create targeted fix/refactor tasks and stop.
• •Pass 2 (after test task Done): verify tests/coverage/priority limits, calculate Quality Score and NFR validation, close Story to Done or create fix tasks.
• •Delegates work to ln-501/ln-502 workers and ln-510-test-planner.

4-Level Gate Model

Verdict calculation: FAIL if any check fails. CONCERNS if minor issues exist. PASS if all clean.

Quality Score

Formula: Quality Score = 100 - (20 × FAIL_count) - (10 × CONCERN_count)

NFR Validation

Evaluate 4 non-functional requirement dimensions:

Additional prefixes: TEST- (coverage gaps), ARCH- (architecture issues), DOC- (documentation gaps)

NFR verdict per dimension: PASS / CONCERNS / FAIL

When to Use

• •Pass 1: all implementation tasks Done; test task missing or not Done.
• •Pass 2: test task exists and is Done.
• •Explicit pass parameter can force 1 or 2; otherwise auto-detect by test task status.

Workflow (concise)

• •Phase 1 Discovery: Auto-discover team/config; select Story; load Story + task metadata (no descriptions), detect test task status.
• •Pass 1 flow (fail fast):
  1. •Invoke ln-501-code-quality-checker. If issues -> create refactor task (Backlog), stop.
  2. •Run all linters from tech_stack.md. If fail -> create lint-fix task, stop.
  3. •Invoke ln-502-regression-checker. If fail -> create regression-fix task, stop.
  4. •Invoke ln-510-test-planner (orchestrates: ln-511-test-researcher → ln-512-manual-tester → ln-513-auto-test-planner). If manual testing fails -> create bug-fix task, stop. If all passed -> test task created/updated.
  5. •If test task exists and Done, jump to Pass 2; if exists but not Done, report status and stop.
• •Pass 2 flow (after test task Done):
  1. •Load Story/test task; read test plan/results and manual testing comment from Pass 1.
  2. •Verify limits and priority: Priority ≤15; E2E 2-5, Integration 0-8, Unit 0-15, total 10-28; tests focus on business logic (no framework/DB/library tests).
  3. •Ensure Priority ≤15 scenarios and Story AC are covered by tests; infra/docs updates present.
  4. •Calculate Quality Score and NFR validation (see formulas above):
     • •Run NFR checks per dimensions table
     • •Assign issue prefixes: SEC-, PERF-, REL-, MNT-, TEST-, ARCH-, DOC-
     • •Calculate Quality Score
  5. •Determine Gate verdict per 4-Level Gate Model above

TodoWrite format (mandatory): Add pass steps to todos before starting:

Pass 1:
- Invoke ln-501-code-quality-checker (in_progress)
- Run linters from tech_stack.md (pending)
- Invoke ln-502-regression-checker (pending)
- Invoke ln-510-test-planner (research + manual + auto tests) (pending)

Pass 2:
- Verify test task coverage (in_progress)
- Mark Story Done (pending)

Mark each as in_progress when starting, completed when done. On failure, mark remaining as skipped.

Worker Invocation (MANDATORY)

CRITICAL: All delegations use Task tool with subagent_type: "general-purpose" for context isolation.

Prompt template:

Task(description: "[Pass 1/2] quality check via ln-50X",
     prompt: "Execute ln-50X-{worker}. Read skill from ln-50X-{worker}/SKILL.md. Story: {storyId}",
     subagent_type: "general-purpose")

Note: ln-510 orchestrates the full test pipeline (ln-511 research → ln-512 manual → ln-513 auto tests).

❌ FORBIDDEN SHORTCUTS (Anti-Patterns):

• •Running mypy, ruff, pytest directly instead of invoking ln-501/ln-502
• •Doing "minimal quality check" (just linters) and skipping ln-510 test planning
• •Asking user "Want me to run the full skill?" after doing partial checks
• •Marking steps as "completed" in todo without invoking the actual skill
• •Any command execution that should be delegated to a worker skill

✅ CORRECT BEHAVIOR:

• •Use Skill(skill: "ln-50X-...") for EVERY step — NO EXCEPTIONS
• •Wait for each skill to complete before proceeding
• •If skill fails → create fix task → STOP (fail fast)
• •Never bypass skills with "I'll just run the command myself"

ZERO TOLERANCE: If you find yourself running quality commands directly (mypy, ruff, pytest, curl) instead of invoking the appropriate skill, STOP and use Skill tool instead.

Critical Rules

• •Early-exit: any failure creates a specific task and stops Pass 1/2.
• •Single source of truth: rely on Linear metadata for tasks; kanban is updated by workers/ln-400.
• •Task creation via skills only (ln-510/ln-301); this skill never edits tasks directly.
• •Pass 2 only runs when test task is Done; otherwise return error/status.
• •Language preservation in comments (EN/RU).

Definition of Done

• •Pass 1: ln-501 pass OR refactor task created; linters pass OR lint-fix task created; ln-502 pass OR regression-fix task created; ln-510 pipeline pass (research + manual + auto tests) OR bug-fix task created; test task created/updated; exits.
• •Pass 2: test task verified (priority/limits/coverage/infra/docs); Quality Score calculated; NFR validation completed; Gate verdict determined (PASS/CONCERNS/FAIL/WAIVED).
• •Gate output format:
  yaml

  gate: PASS | CONCERNS | FAIL | WAIVED
  quality_score: {0-100}
  nfr_validation:
    security: PASS | CONCERNS | FAIL
    performance: PASS | CONCERNS | FAIL
    reliability: PASS | CONCERNS | FAIL
    maintainability: PASS | CONCERNS | FAIL
  issues: [{id: "SEC-001", severity: high|medium|low, finding: "...", action: "..."}]
  

• •Story set to Done (PASS/CONCERNS/WAIVED) or fix tasks created (FAIL); comment with gate verdict posted.

Reference Files

• •Gate levels: references/gate_levels.md (detailed scoring rules and thresholds)
• •Workers: ../ln-501-code-quality-checker/SKILL.md, ../ln-502-regression-checker/SKILL.md
• •Test planning orchestrator: ../ln-510-test-planner/SKILL.md (coordinates ln-511/512/513)
• •Tech stack/linters: docs/project/tech_stack.md

Version: 5.0.0 (Added 4-level Gate Model, Quality Score, NFR validation based on BMAD qa-gate methodology) Last Updated: 2026-01-29