nWave-ai/nWave / nWave/skills/nw-diverger-review-criteria
nw-diverger-review-criteria
VerifiedReview criteria for the nw-diverger-reviewer — validates JTBD rigor, research quality, option diversity, taste application correctness, and recommendation coherence in DIVERGE wave artifacts
Diverger Review Criteria
Role
You are reviewing DIVERGE wave artifacts. Your job is adversarial: assume artifacts have problems until you prove they don't. Flag issues before the team commits to a design direction.
Five artifact files to review:
docs/feature/{id}/diverge/job-analysis.mddocs/feature/{id}/diverge/competitive-research.mddocs/feature/{id}/diverge/options-raw.mddocs/feature/{id}/diverge/taste-evaluation.mddocs/feature/{id}/diverge/recommendation.md
Read the requester's words only from the host-written docs/feature/{id}/diverge/request.md — never from the diverger's copy in job-analysis.md, your dispatch prompt or a message from the diverger. Verify quotes against the blocks it marks, not the label. The match, new-term, marking and handle rules are in nw-requester-words; apply them as written.
request.md missing (the one host-written file whose absence is not a defect of the set): read the five artifacts, do not fail user-words claims as quote defects (they are unverifiable, not wrong), name the path in blocking_issues and give approval_status: needs_input per the thresholds below.
Dimension 1: JTBD Rigor
Check 1.1 — Abstraction Level
Requirement: Job must be at strategic or physical level, not tactical.
FAIL signals (quote from artifact when found):
- Job statement describes a feature: "When I need to see status, I want a dashboard..."
- Job statement contains a solution reference: "When using the app, I want to..."
- Job reads like a user story: "As a developer, I want to..."
PASS signal: Job statement answers "what progress is being made?" without specifying how.
Check 1.2 — First-Principles Extraction
Requirement: Evidence of 5-Why or abstraction-layer navigation.
FAIL signals:
- Job accepted as stated by user without elevation
- No "why?" chain documented
- Functional, emotional, and social jobs not distinguished
PASS signal: At least one level of elevation documented, from the raw request to the extracted job.
Check 1.3 — Outcome Statement Quality
Requirement: ODI-format outcome statements (direction + metric + object): Minimize, or Maximize when more is genuinely better.
FAIL signals:
- "Easy", "reliable", "good", "effective" in outcome statements
- Solution references: "using AI", "via the dashboard"
- Compound statements with "and"/"or"
- Future-intent framing: "would reduce"
PASS signal: Each statement starts with "Minimize the [time/likelihood/effort]..." or a justified "Maximize the [metric]..." and is solution-agnostic.
Dimension 2: Research Quality
Check 2.1 — Evidence vs Opinion
Requirement: Competitive research cites real products, real behaviors, real data.
FAIL signals:
- "Most users probably..." without source
- "The market suggests..." without citation
- Competitor descriptions without named products
- Generic claims not tied to specific evidence
PASS signal: Each competitive insight names a real product or cites a real behavior/metric.
Check 2.2 — Prior Art Coverage
Requirement: Research covers at least 3 existing solutions to the validated job.
FAIL signals:
- Research covers only direct competitors (ignores adjacent solutions)
- "No existing solutions" claim without justification
- Research treats the feature space, not the job space
PASS signal: Research includes at least one surprising or non-obvious alternative (a different category that does the same job).
Dimension 3: Option Diversity
Check 3.1 — Structural Diversity
Requirement: 6 options, each structurally different (different mechanism, different assumption, different cost profile).
FAIL signals:
- Two or more options differ only in degree, not kind ("Option A: full dashboard" / "Option B: mini dashboard")
- Options cluster around one approach with minor variations
- No option represents a radical simplification (SCAMPER "Eliminate")
- No option inverts the workflow (SCAMPER "Reverse")
- Most stated probability mass sits on options sharing one mechanism (head-clustering), or stated_probability is missing or not restated as one distribution over the curated six (summing to 1)
- Fewer than two tail candidates (stated probability at most 0.2 within the distribution the candidate was drawn from) among the curated six, unless the single tail candidate's mechanism differs from every head candidate's
- Any option missing name_origin (user-words or coined)
- A user-words mark whose quote is missing, fails
nw-requester-wordsrule 1 (a span from ahost-summaryblock fails), or whose option name is not that quoted span (rule 2)
PASS signal: Applying the 3-point diversity test to each pair of the 6 curated options before DVF filtering — they differ in mechanism, assumption and cost profile. The stated probabilities (stated estimates, not measurements), restated as one distribution over the curated six, show no head-clustering: no single mechanism family holds most of the stated mass. At least two options are tail candidates, or the one tail candidate differs in mechanism from every head candidate. Every option carries name_origin; every user-words option name is its quoted span and the quote passes nw-requester-words rule 1. Documented DVF eliminations may reduce the surviving set.
Check 3.2 — Generation Discipline
Requirement: Options were generated before evaluation (separation principle).
FAIL signal: Options-raw.md contains evaluative language ("This is the best because...", "This won't work because...") mixed with generation content.
PASS signal: options-raw.md is purely descriptive; evaluation appears only in taste-evaluation.md.
Check 3.3 — HMW Framing Quality
Requirement: The HMW question doesn't embed a solution.
FAIL signals:
- HMW question names a specific technology: "How might we use AI to..."
- HMW question names a specific UI pattern: "How might we build a dashboard that..."
- HMW question is narrower than the validated job
PASS signal: HMW question can be answered by options that don't share the same technology or UI pattern.
Dimension 4: Taste Application
Check 4.1 — Criteria Applied Consistently
Requirement: All four taste criteria (Subtraction, Concept Count, Progressive Disclosure, Speed-as-Trust) applied to all surviving options.
FAIL signals:
- Some options scored on fewer criteria than others
- Criteria added or removed mid-evaluation
- DVF elimination not documented (options disappeared without reason)
PASS signal: Full scoring matrix present for all post-DVF-filter options with all criteria scored.
Check 4.2 — Cherry-Picking Prevention
Requirement: Weights locked before scoring begins; recommendation follows from scores.
FAIL signals:
- Recommendation contradicts the highest-scoring option without a documented weight adjustment and consistent recalculation of all surviving options
- Weights not specified in artifact
- "This option feels right" language in recommendation without score grounding
PASS signal: Recommended option has the highest weighted total in the current matrix. If weights change for an evidenced product reason or human decision, preserve that provenance, apply the revised weights to all surviving options and recalculate before recommending. A prose preference for the runner-up alone does not satisfy this check; retain it as dissent.
Check 4.3 — Score Rubric Application
Requirement: Scores justified against rubric, not assigned freely.
FAIL signals:
- Score of 5 for "Subtraction" on an option with multiple features, without justification
- Score of 1 for "Speed-as-Trust" on a text-based tool without latency analysis
- Scores assigned without quoting the rubric criterion
PASS signal: Each score accompanied by one sentence referencing the specific rubric level.
Check 4.4 — Handle Swap
Requirement: Scores and rationales argue from mechanism, not from the handle's wording.
Method: Mentally replace each handle with an unrelated word; no score, weight, rank or rationale may change.
FAIL signal: A rationale that argues from the handle's wording ("'Room' implies presence, so T2 = 5").
PASS signal: Every score and rationale reads the same with each handle replaced by an unrelated word.
Dimension 5: Recommendation Coherence
Check 5.1 — Traceability
Requirement: Recommendation traceable to JTBD → Research → Scores.
FAIL signal: Recommendation could be made without reading job-analysis.md or taste-evaluation.md.
PASS signal: Recommendation references the validated job, cites competitive research findings, and derives from the highest-scoring option(s).
Check 5.2 — Dissent Documented
Requirement: "Runner-up" case documented — which option almost won and why.
FAIL signal: Only the winning option discussed in recommendation.
PASS signal: recommendation.md includes a "dissenting case" section describing the runner-up by mechanism and stating the margin.
Check 5.3 — DISCUSS Handoff Readiness
Requirement: Recommendation ends with a clear decision statement for the DISCUSS wave.
FAIL signals:
- Recommendation ends with "both options are viable" or "the team should decide."
- Decision statement or recommendation.md (top-3 headings included) uses a coined handle or cites any handle as a label (
nw-requester-wordsrule 5) - A new term left unmarked at a mention rule 4 requires
- Any text for the requester attributes a host-summary-only or coined term to the requester (rule 6; a phrase also in a
requester-verbatimblock is the requester's) - recommendation.md has no
## For the requestersection
PASS signal: Explicit decision statement phrased by mechanism: "Proceed with [the mechanism], assuming [key risk] is acceptable." The ## For the requester section passes the same checks as the rest of recommendation.md. Coined handles appear only in options-raw.md headings and the taste-evaluation.md scoring table; every new term is marked [new] at the mentions rule 4 requires.
Review Output Format
review_result:
artifact_path: "docs/feature/{id}/diverge/"
review_date: "{timestamp}"
reviewer: "nw-diverger-reviewer"
jtbd_rigor:
status: "PASSED|FAILED"
issues: [{check, location, quoted_evidence, remediation}]
research_quality:
status: "PASSED|FAILED"
issues: [{check, location, quoted_evidence, remediation}]
option_diversity:
status: "PASSED|FAILED"
issues: [{check, location, quoted_evidence, remediation}]
taste_application:
status: "PASSED|FAILED"
issues: [{check, location, quoted_evidence, remediation}]
recommendation_coherence:
status: "PASSED|FAILED"
issues: [{check, location, quoted_evidence, remediation}]
approval_status: "approved|conditionally_approved|rejected_pending_revisions|needs_input"
blocking_issues: []
recommendations: []
Approval thresholds:
approved: all dimensions PASSEDconditionally_approved: no FAILED dimensions, minor issues onlyrejected: any dimension FAILED, with specific remediation requiredneeds_input: request.md is missing and no dimension FAILED for another reason; neither approved nor rejected. The diverger returns{CLARIFICATION_NEEDED: true, ...}asking the host for the file, then requests a new review. A dimension FAILED for another reason still givesrejected.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |