/feature-review

name: feature-review
description: Scores backlog items with RICE/WSJF/Kano and files GitHub issues for top candidates. Use when triaging a roadmap or prioritizing features for a sprint.
alwaysApply: false
category: workflow-methodology
tags:
- feature-prioritization
- backlog-triage
- RICE
- WSJF
- Kano
- roadmap
dependencies:
- imbue:scope-guard
tools: []
usage_patterns:
- feature-inventory
- prioritization-scoring
- suggestion-generation
- github-integration
- research-enrichment
complexity: intermediate
estimated_tokens: 3500
modules:
- modules/scoring-framework.md
- modules/classification-system.md
- modules/tradeoff-dimensions.md
- modules/research-enrichment.md
- modules/configuration.md
- modules/multi-metric-evaluation-methodology.md

Verification

Run `make test-feature-review` to verify scoring logic after changes.

Feature Review

Review implemented features and suggest new ones using evidence-based prioritization. Create GitHub issues for accepted suggestions.

Philosophy

Feature decisions rely on data. Every feature involves tradeoffs that require evaluation. This skill uses hybrid RICE+WSJF scoring with Kano classification to prioritize work and generates actionable GitHub issues for accepted suggestions.

When To Use

• Roadmap reviews (sprint planning, quarterly reviews).
• Retrospective evaluations.
• Planning new development cycles.

When NOT To Use

• Emergency bug fixes.
• Simple documentation updates.
• Active implementation (use `scope-guard`).

Quick Start

1. Inventory Current Features

Discover and categorize existing features:

/feature-review --inventory

2. Score and Classify

Evaluate features against the prioritization framework:

/feature-review

3. Generate Suggestions

Review gaps and suggest new features:

/feature-review --suggest

4. Research-Enriched Scoring

Use tome plugin to adjust scores with external evidence:

/feature-review --research

5. Upload to GitHub

Create issues for accepted suggestions:

/feature-review --suggest --create-issues

Workflow

Phase 1: Feature Discovery (`feature-review:inventory-complete`)

Identify features by analyzing:

1. **Code artifacts**: Entry points, public APIs, and configuration surfaces. 2. **Documentation**: README lists, CHANGELOG entries, and user docs. 3. **Git history**: Recent feature commits and branches.

**Output:** Feature inventory table.

Phase 2: Classification (`feature-review:classified`)

Classify each feature along two axes:

**Axis 1: Proactive vs Reactive**

| Type | Definition | Examples | |------|------------|----------| | **Proactive** | Anticipates user needs. | Suggestions, prefetching. | | **Reactive** | Responds to explicit input. | Form handling, click actions. |

**Axis 2: Static vs Dynamic**

| Type | Update Pattern | Storage Model | |------|---------------|---------------| | **Static** | Incremental, versioned. | File-based, cached. | | **Dynamic** | Continuous, streaming. | Database, real-time. |

See [classification-system.md](modules/classification-system.md) for details.

Phase 3: Scoring (`feature-review:scored`)

Apply hybrid RICE+WSJF scoring:

Feature Score = Value Score / Cost Score

Value Score = weighted_avg(Reach, Impact, Business Value, Time Criticality)
Cost Score = weighted_avg(Effort, Risk, Complexity)

Adjusted Score = Feature Score * Confidence

**Scoring Scale:** Fibonacci (1, 2, 3, 5, 8, 13). Default weights (value: reach 0.25, impact 0.30, business value 0.25, time criticality 0.20; cost: effort 0.40, risk 0.30, complexity 0.30) and how to override them are in the scoring framework module.

**Thresholds:**

• **> 2.5**: High priority.
• **1.5 - 2.5**: Medium priority.
• **< 1.5**: Low priority.

See [scoring-framework.md](modules/scoring-framework.md) for the framework. See [multi-metric-evaluation-methodology.md](modules/multi-metric-evaluation-methodology.md) when one model is not enough: it covers how to combine RICE, WSJF, and Kano, where each model fits, and how to reconcile conflicting signals.

Phase 4: Tradeoff Analysis (`feature-review:tradeoffs-analyzed`)

Evaluate each feature across quality dimensions:

| Dimension | Question | Scale | |-----------|----------|-------| | **Quality** | Does it deliver correct results? | 1-5 | | **Latency** | Does it meet timing requirements? | 1-5 | | **Token Usage** | Is it context-efficient? | 1-5 | | **Resource Usage** | Is CPU/memory reasonable? | 1-5 | | **Redundancy** | Does it handle failures gracefully? | 1-5 | | **Readability** | Can others understand it? | 1-5 | | **Scalability** | Will it handle 10x load? | 1-5 | | **Integration** | Does it play well with others? | 1-5 | | **API Surface** | Is it backward compatible? | 1-5 |

See [tradeoff-dimensions.md](modules/tradeoff-dimensions.md) for criteria.

Phase 4.5: Research Enrichment (`feature-review:research-enriched`)

**Triggered by:** `--research` flag. Requires tome plugin.

Use tome's multi-source research to adjust scoring factors with external evidence. This phase runs between tradeoff analysis and gap analysis.

1. **Dispatch research**: For each feature, construct research topics and dispatch tome channels (code-search, discourse, papers, triz) in parallel. 2. **Synthesize findings**: Merge results across channels using `tome:synthesize`. 3. **Calculate deltas**: Map findings to scoring factor adjustments using channel-to-factor mapping. 4. **Apply deltas**: Adjust initial scores by research deltas, clamp to Fibonacci scale, respect max_delta. 5. **Present evidence**: Show adjustment table with evidence sources and rationale.

See [research-enrichment.md](modules/research-enrichment.md) for the full enrichment protocol, delta calculation, and graceful degradation behavior.

**Graceful degradation**: If tome is not installed, prints a warning and proceeds with initial scores unchanged.

Phase 5: Ga