Approach Selection for Claude Prompts

Calibration: Tier 2, Opus-primary. See repository README for model compatibility.

This Skill provides the decision logic for choosing the right identity, reasoning method, and output format when building a Claude prompt. Given a task description, it maps task characteristics to the best-fit approach in each category.

When to Use This Skill

Use this when you need to decide:

• Which identity (role) fits a task
• Which reasoning method produces the best analytical depth for the task shape
• Which output format matches the deliverable
• When to combine approaches across categories
• When core approaches are sufficient vs. when domain-specific approaches add value

This Skill provides selection logic and approach summaries. For the full approach text (ready to paste into a prompt), consult the reference files in references/.

Important

Be decisive. When one approach is clearly the best fit, recommend it — do not present three options and ask the user to choose unless the options lead to genuinely different outcomes.

Match complexity to task. Not every task needs all four selection decisions. A simple task may need only an identity and output format. A medium task adds a reasoning method. Only complex, high-stakes tasks need the full selection plus quality control additions.

Core approaches cover 80% of tasks. The 8 identity approaches, 18 reasoning variants, and 10 output formats handle most prompt construction needs. Route to domain-specific approaches only when the task requires specialization the core catalog cannot provide. See the domain routing guidance at the end of this file.

Reasoning discipline

Before recommending an approach, walk through the task characteristics explicitly. State the observations (task type, depth required, audience, output shape, constraints), name the approach pattern they match against the decision tree, then apply the selection. Do not compress this sequence into a summary recommendation.

If the task scope is unclear (single-prompt selection? multi-block combination? architecture decision?), confirm scope with the user before proceeding. Do not proceed on inferred assumptions.

Step 1: Classify the Task

Before selecting approaches, classify the task using this routing table. When a task matches multiple categories, lead with the dominant one and fold in elements from the secondary.

Strategic / Advisory — Business strategy, market entry, competitive analysis, organizational decisions. Identity: Strategic Advisor or Product Strategist. Reasoning: Market & Competitive Strategy, Resource Allocation, or Change & Transformation. Output: Executive Brief, Strategic Memo, or Decision Matrix.

Technical / Engineering — System design, infrastructure, migration planning, technical evaluation. Identity: Technical Architect. Reasoning: System Design, Debugging & Incident Analysis, or Migration & Transition. Output: Technical Design Document or Implementation Plan.

Analytical / Evaluative — Evidence review, data analysis, root cause investigation, risk evaluation. Identity: Research Synthesist or Financial Analyst. Reasoning: General Analysis, Root Cause Diagnosis, or Risk Assessment. Output: Research Summary, Post-Mortem, or Decision Matrix.

Creative / Generative — Concept development, messaging, content creation, narrative design. Identity: Communications Strategist or custom. Reasoning: Concept Development, Messaging & Narrative, or Solution Ideation. Output: Custom, matched to deliverable type.

Research / Synthesis — Literature review, evidence integration, landscape overview. Identity: Research Synthesist. Reasoning: Evidence Synthesis, Landscape Scan, or Gap Analysis. Output: Research Summary or Competitive Analysis.

Comparative / Decision — Option evaluation, vendor selection, prioritization of a set. Identity: Matched to domain. Reasoning: Option Evaluation, Vendor/Tool Selection, or Prioritization. Output: Decision Matrix or Executive Brief.

Operational / Process — Process design, workflow optimization, operational systems. Identity: Operations Designer. Reasoning: General Analysis or custom. Output: Process Documentation, Implementation Plan, or Stakeholder Update.

Educational / Explanatory — Teaching, explaining, training material, progressive understanding. Identity: Educator/Explainer. Reasoning: Light — focus on progressive complexity. Output: Custom, matched to format (guide, explainer, training material).

Step 2: Select Identity

The identity determines WHO Claude is for the task — shaping depth, vocabulary, reasoning style, and what Claude treats as obvious vs. requiring explanation.

For full approach text and failure modes, see references/identity-blocks.md.

Identity Decision Tree

1. What domain is the task in?
   → Business strategy, market analysis, competitive dynamics → Strategic Advisor
   → System design, infrastructure, technical evaluation → Technical Architect
   → Evidence review, data analysis, research synthesis → Research Synthesist
   → Process design, workflow, operational systems → Operations Designer
   → Financial analysis, valuation, investment → Financial Analyst
   → Product decisions, roadmap, feature evaluation → Product Strategist
   → Messaging, positioning, narrative, content → Communications Strategist
   → Teaching, explaining, training material → Educator / Explainer

2. Does the task sit at an intersection of two domains?
   → Yes → Build a hybrid combining the two relevant approaches.
   → No → Use the single best-fit approach.

3. Does the task require a different seniority calibration?
   → Expert audience → Use "principal" or "world-class" — more nuanced,
     assumption-challenging output.
   → Audience needs step-by-step guidance → Reduce seniority — more
     explanatory output.
   → Default → "Senior" is right for most tasks.

4. Does no approach fit at all?
   → Build custom using the template:
     [SENIORITY] [ROLE] with expertise in [DOMAINS].
     Approaches problems by [REASONING STYLE].
     Prioritizes [VALUE 1] over [VALUE 2].
     [ONE BEHAVIORAL SENTENCE].

Identity Calibration Principles

Seniority shapes the output. "Senior" or "principal" produces nuanced analysis with tradeoffs acknowledged. Removing seniority produces more explanatory, step-by-step output. Match seniority to the depth you need.

Domain intersections create distinctive thinking. "Data scientist with supply chain expertise" produces different analysis than either role alone.

Stated values resolve ambiguity. When Claude faces a tradeoff (thoroughness vs. speed, precision vs. accessibility), the values statement tells it which way to lean.

The behavioral sentence prevents drift. One concrete instruction about communication style anchors the identity more than additional domain descriptions.

Step 3: Select Reasoning

The reasoning method controls HOW Claude thinks through the problem — the analytical steps, order of operations, and where attention is directed. This is the highest-leverage selection: the difference between shallow and deep output almost always traces back to reasoning quality.

For full approach text and failure modes, see references/reasoning-blocks.md.

Reasoning Decision Tree

1. What is the primary task shape?
   → Evaluate something that exists → Analytical family
   → Plan or decide a strategic direction → Strategic family
   → Generate something new → Creative family
   → Build or fix a technical system → Technical family
   → Synthesize information across sources → Research family
   → Compare options and recommend → Comparative family

2. Within the family, which variant?

   Analytical:
   → General evaluation of a situation → General Analysis
   → Something is failing, find out why → Root Cause Diagnosis
   → Evaluate risks of a decision → Risk Assessment

   Strategic:
   → Market opportunity or