Nw Ad Critique DimensionsSAFE

AI agents that guide you from idea to working code, with you in control at every step.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: nw-ad-critique-dimensions
description: Review dimensions for acceptance test quality - happy path bias, GWT compliance, business language purity, coverage completeness, walking skeleton user-centricity, priority validation, observable behavior assertions, traceability coverage, and walking skeleton boundary proof
user-invocable: false
disable-model-invocation: true
---

# Acceptance Test Critique Dimensions

Load when performing peer review of acceptance tests (during *handoff-develop).

## Dimension 1: Happy Path Bias

**Pattern**: Only successful scenarios, error paths missing.

Detection: Count success vs error scenarios. Error should be at least 40%. Missing coverage examples: login success but no invalid password | Payment processed but no decline/timeout | Search results but no empty/error cases.

Severity: blocker (production error handling untested).

## Dimension 2: GWT Format Compliance

**Pattern**: Scenarios violate Given-When-Then structure.

Violations: Missing Given context | Multiple When actions (split into separate scenarios) | Then with technical assertions instead of business outcomes. Each scenario: Given (context), When (single action), Then (observable outcome).

Severity: high (tests not behavior-driven).

## Dimension 3: Business Language Purity

**Pattern**: Technical terms leak into acceptance tests.

Flag: database, API, HTTP, REST, JSON, classes, methods, services, controllers, status codes (500, 404), infrastructure (Redis, Kafka, Lambda).

Business alternatives: "Customer data is stored" not "Database persists record" | "Order is confirmed" not "API returns 200 OK" | "Payment fails" not "Gateway throws exception"

Severity: high (tests coupled to implementation).

## Dimension 4: Coverage Completeness

**Pattern**: User stories lack acceptance test coverage.

Validation: Map each story to scenarios | Verify all AC have corresponding tests | Confirm edge cases and boundaries tested.

Severity: blocker (unverified requirements).

## Dimension 5: Walking Skeleton User-Centricity

**Pattern**: Walking skeletons describe technical layer connectivity instead of user value.

Detection litmus test for `@walking_skeleton` scenarios:
- Title describes user goal or technical flow?
- Then steps describe user observations or internal side effects?
- Could non-technical stakeholder confirm "yes, that is what users need"?

Violations: "End-to-end order flow through all layers" (technical framing) | Then "order row inserted in database" (internal side effects) | Given "database contains user record" instead of "customer has an account"

Severity: high (skeletons that only prove wiring miss the point -- first skeleton should be demo-able to stakeholder).

## Dimension 6: Priority Validation

**Pattern**: Tests address secondary concerns while larger gaps exist.

Questions: 1. Is this the largest bottleneck? (timing data or gap analysis) | 2. Simpler alternatives considered? | 3. Constraint prioritization correct? | 4. Test design decisions data-justified?

Severity: blocker if wrong problem addressed, high if no measurement data.

## Dimension 7: Observable Behavior Assertions

**Pattern**: Tests assert internal state or method calls instead of observable behavior.

For EVERY Then step in EVERY scenario, apply this mechanical checklist:

1. Does the assertion check a return value from a driving port call? YES = pass, NO = flag.
2. Does the assertion check an observable outcome (user sees X, system produces Y)? YES = pass, NO = flag.
3. Does the assertion check internal state, private fields, or method call counts? YES = REJECT the scenario.

**Concrete violations to flag**:
- `assert mock_repo.save.called` — asserts method call, not observable outcome
- `assert len(db.query(Order).all()) == 1` — asserts internal DB state
- `assert obj._internal_field == "value"` — asserts private state
- `assert os.path.exists("output.json")` — asserts file existence (implementation detail)

**Concrete passing assertions**:
- `assert result.is_confirmed()` — observable business outcome
- `assert result.order_number is not None` — return value from driving port
- `assert "confirmation" in customer_notification.subject` — observable user outcome

**Relationship to Dim 5 (Walking Skeleton User-Centricity)**:
- Dim 5 validates walking skeleton SCOPE (user goal framing vs technical layer framing)
- Dim 7 validates ASSERTION TYPE for ALL scenarios (walking skeletons AND focused scenarios)
- A scenario can pass Dim 5 (good user-centric framing) and fail Dim 7 (internal state assertions)

Severity: high (tests coupled to implementation break on refactoring).

## Dimension 8: Traceability Coverage

**Pattern**: Scenarios exist without traceability to upstream wave artifacts.

Two mandatory traceability checks:

**Check A — Story-to-Scenario mapping**:
1. Read `docs/feature/{feature-id}/discuss/user-stories.md`
2. Extract ALL story IDs (e.g., US-01, US-02)
3. For EACH story ID, verify at least one scenario references it (via tag or comment)
4. Flag EVERY story ID with zero matching scenarios as BLOCKER

**Check B — Environment-to-Scenario mapping**:
1. Read `docs/feature/{feature-id}/devops/environments.yaml`
2. If missing, use defaults: `clean`, `with-pre-commit`, `with-stale-config`
3. For EACH environment, verify at least one walking skeleton includes a Given clause referencing that environment's preconditions
4. Flag EVERY environment with zero matching Given clauses as HIGH

**What this dimension does NOT cover**:
- KPI measurability — that is PO-reviewer scope during DELIVER post-merge gate
- Scenario quality — covered by Dims 1-7

Severity: blocker for Check A (untraceable requirements), high for Check B (untested environments).

## Review Output Format

```yaml
review_id: "accept_rev_{timestamp}"
reviewer: "acceptance-designer (review mode)"

strengths:
  - "{positive test design aspect with example}"

issues_identified:
  happy_path_bias:
    - issue: "Feature {name} only tests suc

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.