feature-evaluator

Files (3)

• SKILL.md
• context-template.md · references/
• evaluation-criteria.md · references/

SKILL.md (first lines)

# Feature Evaluator

Evaluate proposed features through structured critical analysis to determine value, alternatives, and optimal implementation paths.

## Execution Modes

This skill adapts to the pipeline execution mode:

| Mode | Behavior |
|------|----------|
| `full` | All phases, full depth |
| `fast` | Combined eval+plan output, risk-based gating |
| `prototype` | Abbreviated output (value + risks only) |
| `spike` | Skip evaluation entirely |

## Adaptive Depth

Scale evaluation depth to feature stakes:

| Feature Scale | Phases to Complete | Time Investment |
|---------------|-------------------|-----------------|
| **Small** (< 1 week, easily reversible) | Phases 1, 2, 5 (abbreviated) | 5-10 min |
| **Medium** (1-4 weeks, moderate risk) | All phases, standard depth | 15-30 min |
| **Large** (> 1 month, high stakes/irreversible) | All phases + reference file deep-dive | 30-60 min |

Ask about scale if unclear. Default to Medium.

## Context Gathering

If context is missing, request these essentials before evaluating:

**Must have:**
- What the project does (1-2 sentences)
- What the proposed feature does (plain English)
- Who requested it and why

**Helpful for deeper analysis:**
- Current pain points driving the request
- Existing components that might be relevant
- Constraints (deadline, team size, risk tolerance)
- Success metrics that matter

See [references/context-template.md](references/context-template.md) for full context template when doing formal evaluations.

## Evaluation Process

Execute phases in order. Produce concrete, specific outputs — avoid generic statements.

### Phase 1: Value Articulation

Analyze the feature's value proposition:

```
VALUE PROPOSITION
─────────────────
Problem: [Specific problem in one sentence]
User value: [What gets easier/faster/safer]
Business value: [Cost saved, risk reduced, revenue enabled]
Who benefits: [Specific user segments]
Who doesn't: [Users unaffected or potentially harmed]

Status quo cost: [What happens if we do nothing — quantify if possible]

Strategic fit: [High/Medium/Low] — [Why]
Urgency: [High/Medium/Low] — [Why now vs. later]

RISK ASSESSMENT
───────────────
Estimated effort: [< 1 week / 1-4 weeks / > 1 month]
Reversibility: [Easy / Moderate / Difficult / Irreversible]
Scope clarity: [Clear / Moderate / Unclear]

Risk tier: [Low / Medium / High]
```

**Risk Tier Determination:**
- **Low:** Effort < 1 week AND easily reversible AND clear scope
- **Medium:** Any moderate factor, no high-risk factors
- **High:** Effort > 1 month OR irreversible OR unclear scope

Risk tier controls flow in `fast` and `prototype` modes.
User value: [What gets easier/faster/safer]
Business value: [Cost saved, risk reduced, revenue enabled]
Who benefits: [Specific user segments]
Who doesn't: [Users unaffected or potentially harmed]

Status quo cost: [What happens if we do nothing — quantify if possible]

Strategic fit: [High/Medium/Low] — [Why]
Urgency: [High/Medium/Low] — [Why now vs. later]
```

The "status quo cost" is critical — it forces honest ROI assessment.

### Phase 2: Critical Analysis

Systematically challenge the feature:

**Why this might NOT help:**
- Misalignment risks (wrong problem, rare edge case, solves symptom not cause)
- Adoption risks (discoverability, habit change, trust)
- Complexity/maintenance burden
- Performance/reliability risks
- Security/privacy/compliance risks

**Alternative approaches using existing functionality:**
- What already exists that could be combined/reconfigured?
- Would documentation, training, or better defaults solve this?
- Is there a third-party solution?
- Would a different feature solve this AND other problems?

```
CRITICAL ANALYSIS
─────────────────
Potential drawbacks:
• [Concern 1] — Severity: [High/Med/Low], Likelihood: [High/Med/Low]
• [Concern 2] — Severity: [High/Med/Low], Likelihood: [High/Med/Low]
• [Concern 3] — Severity: [High/Med/Low], Likelihood: [High/Med/Low]

From autoencoders/skills, CC-BY-4.0.

More from autoencoders/skills