tools-only Evolution Scoring Framework
The Evolution/Timelessness lens is the core evaluative perspective for SkillForge v4.0. Every skill must score ≥7 on timelessness to be approved.
Overview
Evolution scoring answers: "How will this skill age?"
Skills that score poorly become:
- Maintenance burdens
- Sources of confusion as they become outdated
- Obstacles to ecosystem evolution
- Technical debt for future Claude users
Requirement: All generated skills must score ≥7/10 on timelessness.
Timelessness Scoring Rubric
Score 1-2: Ephemeral
Characteristics:
- Tightly coupled to specific API versions
- Hardcoded tool versions or paths
- Addresses a transient trend
- No consideration of change
Example:
# Skill: Format for GPT-3.5-Turbo
Formats prompts specifically for the gpt-3.5-turbo-0301 API...
Lifespan: Weeks to months
Verdict: Reject. Fundamentally flawed approach.
Score 3-4: Short-Lived
Characteristics:
- Depends on specific tooling likely to change
- Addresses problem in way specific to current moment
- Some structure but no extension consideration
- May reference volatile external resources
Example:
# Skill: React Class Component Generator
Generates class components following React 17 patterns...
Lifespan: 6 months to 1 year
Verdict: Reject. Needs fundamental redesign with evolution in mind.
Score 5-6: Moderate Longevity
Characteristics:
- Sound core approach
- Some hardcoded elements that may change
- Missing explicit extension points
- Limited temporal projection
Example:
# Skill: API Documentation Generator
Generates OpenAPI 3.0 specifications...
(No consideration for OpenAPI 4.0 or alternatives)
Lifespan: 1-2 years
Verdict: Requires revision. Add extension points, document evolution path.
Score 7-8: Solid Design
Characteristics:
- Principle-based rather than implementation-specific
- Clear extension points documented
- Dependencies abstracted appropriately
- Temporal projection completed
- Graceful degradation designed
Example:
# Skill: API Documentation Generator
Generates API documentation following industry standards.
Supports OpenAPI 3.x with extension point for future specs.
Pattern-based approach adaptable to new documentation formats.
Lifespan: 2-4 years
Verdict: Approved. Good balance of current utility and future resilience.
Score 9-10: Timeless
Characteristics:
- Addresses fundamental, unchanging problem
- Completely principle-based
- Highly composable with other skills
- Designed for evolution from day one
- Could serve as a pattern for other skills
Example:
# Skill: Systematic Problem Decomposition
Breaks complex problems into verifiable atomic steps.
Pattern: Define → Decompose → Verify → Iterate
(This process is timeless regardless of technology)
Lifespan: 5+ years
Verdict: Exemplary. Consider as template for other skills.
Scoring Process
Step 1: Temporal Projection Analysis
For each time horizon, evaluate:
## Temporal Projection
### 6 Months
- **Usage patterns:** How will typical usage differ?
- **Ecosystem changes:** What in the environment might change?
- **Risk assessment:** What could break?
- **Confidence:** High/Medium/Low
### 1 Year
- **Usage patterns:** ...
- **Ecosystem changes:** ...
- **Risk assessment:** ...
- **Confidence:** ...
### 2 Years
- **Usage patterns:** ...
- **Ecosystem changes:** ...
- **Risk assessment:** ...
- **Confidence:** ...
### 5 Years
- **Core problem relevance:** Is the problem still relevant?
- **Approach validity:** Is the solution approach still valid?
- **Confidence:** ...
Step 2: Dependency Stability Assessment
Classify all dependencies:
| Dependency | Type | Stability | Change Likelihood | Abstraction |
|---|---|---|---|---|
| Claude API | External | Stable | Low (backward compat) | Direct |
| Specific model ID | External | Volatile | High | Should abstract |
| Skill patterns | Internal | Stable | Low | Direct OK |
| Specific tool | External | Moderate | Medium | Should abstract |
Abstraction requirement: Volatile and Moderate-stability dependencies MUST be abstracted.
Step 3: Extension Point Evaluation
For each extension point:
| Extension Point | Type | Use Case | Mechanism |
|----------------|------|----------|-----------|
| Pattern plugins | Feature | New architecture patterns | Add to patterns/ |
| Lens additions | Capability | New thinking models | Add to lenses/ |
| Output formats | Output | New file formats | Template-based |
| Validation rules | Quality | New checks | Config-based |
Minimum requirement: At least 2 documented extension points.
Step 4: Anti-Obsolescence Pattern Check
Verify these patterns are applied:
| Pattern | Applied? | Evidence |
|---|---|---|
| Principle over implementation | ☐ | Shows general approach, not specific tool |
| WHY documented | ☐ | Decisions include rationale |
| Loose coupling | ☐ | Dependencies abstracted |
| Graceful degradation | ☐ | Fallbacks for breaking changes |
| Version-agnostic | ☐ | No hardcoded versions |
| Ecosystem-aware | ☐ | Considers related skills |
Common Evolution Anti-Patterns
Anti-Pattern 1: Version Pinning
Wrong:
Uses claude-3-5-sonnet-20241022 for analysis...
Right:
Uses configured model (default: claude-opus-4-6, configurable)...
Anti-Pattern 2: Tool-Specific Design
Wrong:
# ESLint Config Generator
Generates .eslintrc.json for ESLint 8...
Right:
# Linting Configuration Generator
Generates linting configuration for JavaScript/TypeScript.
Default: ESLint (configurable via lint_tool parameter).
Extension point for future linting tools.
Anti-Pattern 3: Missing Extension Points
Wrong:
## Supported Patterns
- Single-Phase
- Multi-Phase
- Orchestrator
(closed list)
Right:
## Supported Patterns
Built-in patterns: Single-Phase, Multi-Phase, Orchestrator
Custom patterns: Add to `patterns/` directory following template
Anti-Pattern 4: Implicit Dependencies
Wrong:
Run `npm run build` to verify...
Right:
Run verification command (default: `npm run build`,
configurable via BUILD_COMMAND or skill config)
Anti-Pattern 5: Point-in-Time Assumptions
Wrong:
Since Claude now supports 200K context...
Right:
Leverage available context window
(designed to scale with future context increases)
Score Adjustment Factors
Positive Adjustments (+0.5 to +1)
| Factor | Adjustment | Criteria |
|---|---|---|
| Exemplary extension design | +1 | Clear, documented, tested extension mechanism |
| Principle documentation | +0.5 | WHY documented for all decisions |
| Ecosystem integration | +0.5 | Seamless composition with 3+ skills |
| Graceful degradation | +0.5 | Handles all identified breaking changes |
Negative Adjustments (-0.5 to -2)
| Factor | Adjustment | Criteria |
|---|---|---|
| Hardcoded version | -1 | Any non-abstracted version reference |
| Missing extension points | -1 | Fewer than 2 documented |
| No temporal projection | -2 | No 6mo/1yr/2yr analysis |
| Tight external coupling | -1 | Direct dependency on volatile resource |
Evolution Score Template
## Evolution Analysis for [Skill Name]
### Base Score Assessment
| Criterion | Score (1-10) | Evidence |
|-----------|--------------|----------|
| Core approach timelessness | X | [evidence] |
| Dependency management | X | [evidence] |
| Extension point quality | X | [evidence] |
| Temporal projection depth | X | [evidence] |
| **Average** | **X.X** | |
### Adjustments
| Factor | Adjustment | Reason |
|--------|------------|--------|
| [Factor] | +/-X | [Reason] |
### Final Score: X/10
### Verdict: APPROVED / NEEDS REVISION
### If Needs Revision:
1. [Specific improvement required]
2. [Specific improvement required]
Integration with Synthesis Panel
The Evolution/Timelessness Agent in the synthesis panel uses this framework to evaluate generated skills. The agent:
- Applies the full scoring rubric
- Documents score justification
- Identifies specific improvements if score < 7
- Only approves if score ≥ 7
Unanimous approval requires all three agents, including Evolution Agent, to approve.