Back to Skills

negative-contrastive-framing

lyndonkl
Updated Today
19 views
5
5
View on GitHub
Metaaidesign

About

This Claude Skill helps define concepts by contrasting them with what they are not, using counterexamples and anti-patterns. It is ideal for clarifying ambiguous boundaries, preventing common mistakes, and refining requirements. Developers should use it when discussing near-miss examples, anti-goals, or what not to do.

Documentation

Negative Contrastive Framing

Table of Contents

Purpose

Define concepts, quality criteria, and boundaries by showing what they're NOT—using anti-goals, near-miss examples, and failure patterns to create crisp decision criteria where positive definitions alone are ambiguous.

When to Use

Clarifying Fuzzy Boundaries:

  • Positive definition exists but edges are unclear
  • Multiple interpretations cause confusion
  • Team debates what "counts" as meeting criteria
  • Need to distinguish similar concepts

Teaching & Communication:

  • Explaining concepts to learners who need counterexamples
  • Training teams to recognize anti-patterns
  • Creating style guides with do's and don'ts
  • Onboarding with common mistake prevention

Setting Standards:

  • Defining code quality (show bad patterns)
  • Establishing design principles (show violations)
  • Creating evaluation rubrics (clarify failure modes)
  • Building decision criteria (identify disqualifiers)

Preventing Errors:

  • Near-miss incidents revealing risk patterns
  • Common mistakes that need explicit guards
  • Edge cases that almost pass but shouldn't
  • Subtle failures that look like successes

What Is It

Negative contrastive framing defines something by showing what it's NOT:

Types of Negative Examples:

  1. Anti-goals: Opposite of desired outcome ("not slow" → define fast)
  2. Near-misses: Examples that almost qualify but fail on key dimension
  3. Failure patterns: Common mistakes that violate criteria
  4. Boundary cases: Edge examples clarifying where line is drawn

Example: Defining "good UX":

  • Positive: "Intuitive, efficient, delightful"
  • Negative contrast:
    • ❌ Near-miss: Fast but confusing (speed without clarity)
    • ❌ Anti-pattern: Dark patterns (manipulative design)
    • ❌ Failure: Requires manual to understand basic tasks

Workflow

Copy this checklist and track your progress:

Negative Contrastive Framing Progress:
- [ ] Step 1: Define positive concept
- [ ] Step 2: Identify negative examples
- [ ] Step 3: Analyze contrasts
- [ ] Step 4: Validate quality
- [ ] Step 5: Deliver framework

Step 1: Define positive concept

Start with initial positive definition, identify why it's ambiguous or fuzzy (multiple interpretations, edge cases unclear), and clarify purpose (teaching, decision-making, quality control). See Common Patterns for typical applications.

Step 2: Identify negative examples

For simple cases with clear anti-patterns → Use resources/template.md to structure anti-goals, near-misses, and failure patterns. For complex cases with subtle boundaries → Study resources/methodology.md for techniques like contrast matrices and boundary mapping.

Step 3: Analyze contrasts

Create negative-contrastive-framing.md with: positive definition, 3-5 anti-goals, 5-10 near-miss examples with explanations, common failure patterns, clear decision criteria ("passes if..." / "fails if..."), and boundary cases. Ensure contrasts reveal the why behind criteria.

Step 4: Validate quality

Self-assess using resources/evaluators/rubric_negative_contrastive_framing.json. Check: negative examples span the boundary space, near-misses are genuinely close calls, contrasts clarify criteria better than positive definition alone, failure patterns are actionable guards. Minimum standard: Average score ≥ 3.5.

Step 5: Deliver framework

Present completed framework with positive definition sharpened by negatives, most instructive near-misses highlighted, decision criteria operationalized as checklist, common mistakes identified for prevention.

Common Patterns

By Domain

Engineering (Code Quality):

  • Positive: "Maintainable code"
  • Negative: God objects, tight coupling, unclear names, magic numbers, exception swallowing
  • Near-miss: Well-commented spaghetti code (documentation without structure)

Design (UX):

  • Positive: "Intuitive interface"
  • Negative: Hidden actions, inconsistent patterns, cryptic error messages
  • Near-miss: Beautiful but unusable (form over function)

Communication (Clear Writing):

  • Positive: "Clear documentation"
  • Negative: Jargon-heavy, assuming context, no examples, passive voice
  • Near-miss: Technically accurate but incomprehensible to target audience

Strategy (Market Positioning):

  • Positive: "Premium brand"
  • Negative: Overpriced without differentiation, luxury signaling without substance
  • Near-miss: High price without service quality to match

By Application

Teaching:

  • Show common mistakes students make
  • Provide near-miss solutions revealing misconceptions
  • Identify "looks right but is wrong" patterns

Decision Criteria:

  • Define disqualifiers (automatic rejection criteria)
  • Show edge cases that almost pass
  • Clarify ambiguous middle ground

Quality Control:

  • Identify anti-patterns to avoid
  • Show subtle defects that might pass inspection
  • Define clear pass/fail boundaries

Guardrails

Near-Miss Selection:

  • Near-misses must be genuinely close to positive examples
  • Should reveal specific dimension that fails (not globally bad)
  • Avoid trivial failures—focus on subtle distinctions

Contrast Quality:

  • Explain why each negative example fails
  • Show what dimension violates criteria
  • Make contrasts instructive, not just lists

Completeness:

  • Cover failure modes across key dimensions
  • Don't cherry-pick—include hard-to-classify cases
  • Show spectrum from clear pass to clear fail

Actionability:

  • Translate insights into decision rules
  • Provide guards/checks to prevent failures
  • Make criteria operationally testable

Avoid:

  • Strawman negatives (unrealistically bad examples)
  • Negatives without explanation (show what's wrong and why)
  • Missing the "close call" zone (all examples clearly pass or fail)

Quick Reference

Resources:

  • resources/template.md - Structured format for anti-goals, near-misses, failure patterns
  • resources/methodology.md - Advanced techniques (contrast matrices, boundary mapping, failure taxonomies)
  • resources/evaluators/rubric_negative_contrastive_framing.json - Quality criteria

Output: negative-contrastive-framing.md with positive definition, anti-goals, near-misses with analysis, failure patterns, decision criteria

Success Criteria:

  • Negative examples span boundary space (not just extremes)
  • Near-misses are instructive close calls
  • Contrasts clarify ambiguous criteria
  • Failure patterns are actionable guards
  • Decision criteria operationalized
  • Score ≥ 3.5 on rubric

Quick Decisions:

  • Clear anti-patterns? → Template only
  • Subtle boundaries? → Use methodology for contrast matrices
  • Teaching application? → Emphasize near-misses revealing misconceptions
  • Quality control? → Focus on failure pattern taxonomy

Common Mistakes:

  1. Only showing extreme negatives (not instructive near-misses)
  2. Lists without analysis (not explaining why examples fail)
  3. Cherry-picking easy cases (avoiding hard boundary calls)
  4. Strawman negatives (unrealistically bad)
  5. No operationalization (criteria remain fuzzy despite contrasts)

Key Insight: Negative examples are most valuable when they're almost positive—close calls that force articulation of subtle criteria invisible in positive definition alone.

Quick Install

/plugin add https://github.com/lyndonkl/claude/tree/main/negative-contrastive-framing

Copy and paste this command in Claude Code to install this skill

GitHub 仓库

lyndonkl/claude
Path: skills/negative-contrastive-framing

Related Skills

sglang

Meta

SGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.

View skill

evaluating-llms-harness

Testing

This Claude Skill runs the lm-evaluation-harness to benchmark LLMs across 60+ standardized academic tasks like MMLU and GSM8K. It's designed for developers to compare model quality, track training progress, or report academic results. The tool supports various backends including HuggingFace and vLLM models.

View skill

llamaguard

Other

LlamaGuard is Meta's 7-8B parameter model for moderating LLM inputs and outputs across six safety categories like violence and hate speech. It offers 94-95% accuracy and can be deployed using vLLM, Hugging Face, or Amazon SageMaker. Use this skill to easily integrate content filtering and safety guardrails into your AI applications.

View skill

langchain

Meta

LangChain is a framework for building LLM applications using agents, chains, and RAG pipelines. It supports multiple LLM providers, offers 500+ integrations, and includes features like tool calling and memory management. Use it for rapid prototyping and deploying production systems like chatbots, autonomous agents, and question-answering services.

View skill