skills/topic-thinking/SKILL.md
Manages thinking patterns and mental models when switching between different topics or modes. Use when (1) switching from analysis to design, (2) switching from exploration to validation, (3) switching from reading to writing, (4) any topic transition where thinking approach must change, (5) user signals "we're going to sketch/design/build" after analysis phase.
npx skillsauth add emliunix/home.conf topic-thinkingInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
When switching topics, thinking patterns must reset. This skill ensures appropriate mental models are activated for each mode.
Users signal topic switches with phrases like:
Action on signal: Acknowledge the switch and activate the appropriate thinking mode.
Trigger: Reading/exploring existing code, documentation, or systems. Mental Model: Detective - gathering evidence, tracing paths, understanding causality. Key Questions:
Must Do:
Must Not Do:
Trigger: Creating new architecture, APIs, or systems after analysis. Mental Model: Architect - making tradeoffs, establishing principles, defining boundaries. Key Questions:
Must Do:
Must Not Do:
Trigger: Verifying claims, checking evidence, or testing hypotheses. Mental Model: Skeptic - questioning assumptions, seeking disproof, measuring confidence. Key Questions:
Must Do:
Must Not Do:
Trigger: Writing concrete code, tests, or specifications. Mental Model: Craftsman - precision, completeness, testability. Key Questions:
Must Do:
Must Not Do:
When user signals a topic switch:
Every mode switch MUST be echoed explicitly. This makes the behavior change transparent.
Template:
"[Switch signal detected]. Activating [Mode] ([Mental Model]).
Switching from [Old Mode] to [New Mode] - [Context/Goal]."
Required Elements:
Analysis → Design:
User: "And we're going to sketch our own module system design"
Response: "Switch signal detected: 'we're going to sketch our own
module system design'. Activating Design Mode (Architect mental model).
Switching from Analysis to Design - moving from understanding GHC's
implementation to designing a simplified alternative focused on single
unified environments and explicit loading."
Then proceed with design work...
Design → Implementation:
User: "Let's implement the NameEnv"
Response: "Switch signal detected: 'let's implement'. Activating
Implementation Mode (Craftsman mental model). Switching from Design
to Implementation - moving from architecture to concrete code with
complete type signatures and error handling."
Then proceed with implementation...
Exploration → Validation:
User: "Validate those claims against the source"
Response: "Switch signal detected: 'validate those claims'. Activating
Validation Mode (Skeptic mental model). Switching from Exploration to
Validation - moving from discovery to verification, checking each claim
against actual source code."
Then proceed with validation...
Why Echo:
These apply in ALL modes:
For deep analysis of complex codebases, apply the Tiered Analysis Framework (see TIERED_ANALYSIS_FRAMEWORK.md):
| Tier | Purpose | When to Use | |------|---------|-------------| | Base Theme | Define 3-5 analytical dimensions | Before analyzing code | | Main Theme | Map component relationships | After establishing dimensions | | Detailed Facts | Analyze functions through dimensions | Throughout exploration |
Key insight: Mode switching in this skill corresponds to moving between tiers:
Correct Architecture Over Convenience: When choosing between implementation approaches, prioritize structural correctness over implementation effort.
Key Principles:
ValBind vs anonymous tuple)Anti-Pattern: Effort-Based Decision Making Choosing the "easier" path (fewer files to touch, less code to write) results in technical debt that wastes more effort later. Effort-based shortcuts accumulate into unmaintainable systems.
Correct Approach:
development
Systematic codebase investigation producing structured exploration files (Notes → Facts → Claims). Use when researching unknown systems, tracing code paths, or documenting architecture.
testing
Collaborative design workflow for architecture and system design. Use when the user wants to design or redesign a system, component, or feature. Triggers on phrases like "design", "architecture", "how should we", "what's the best way to", or when the user asks for tradeoff analysis. This skill is for exploration and decision-making, not implementation.
development
Change plan workflow for non-trivial code changes. Create, review, and track changes before implementation.
testing
Skill from parent directory