Skip to content

chaseoneillrnd/prompts

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 

Repository files navigation

Claude AI Skills

A collection of skills for use with Claude Code and Cline to supercharge codebase analysis and implementation planning.


Skills Overview

Skill Purpose When to Use
trace Systematically analyze existing code Before refactoring, debugging, or understanding unfamiliar code
technical-guide-architect Create agent-executable implementation guides Before building new features or systems

🔍 Deep-Trace

File: skills/trace/SKILL.md

Systematically map and analyze execution paths through any codebase using a 7-phase iterative framework. Produces comprehensive documentation in ai_docs/ that makes refactoring and debugging decisions trivial.

What It Does

Transforms "I don't understand this code" into deep, documented understanding by tracing:

  • Entry points and exit points
  • Happy paths, error paths, and edge cases
  • Anti-patterns, bottlenecks, and architectural issues
  • Visual Mermaid diagrams of execution flows

Usage

Claude Code

@deep-trace <feature or execution path to analyze>

Examples:

@deep-trace the user authentication flow and identify security gaps
@deep-trace how orders are processed from cart to fulfillment
@deep-trace the WebSocket connection lifecycle and find memory leaks

Cline

Add the skill content to your system prompt or reference it in your task:

Using the Deep-Trace methodology, analyze <feature> through all 7 phases.
Start with Phase 1 (Lightning Overview) and pause for my approval before continuing.

The 7 Phases

Phase Name Output
1 Lightning Overview & Path Mapping ai_docs/notes/phase1-overview.md
2 Path Analysis & Alternative Assessment ai_docs/notes/phase2-analysis.md
3 Deep Execution Tracing ai_docs/notes/phase3-deep-trace.md
4 Pattern Analysis & Issue Identification ai_docs/notes/phase4-issues.md
5 Solution Synthesis & Documentation ai_docs/solutions/
6 Visualization & Architecture Mapping ai_docs/diagrams/
7 Final Review & Deliverable Packaging Executive summary + roadmap

Tip: The agent pauses after each phase for your approval. You can stop early — every phase delivers standalone value.

Output Structure

ai_docs/
├── notes/
│   ├── phase1-overview.md
│   ├── phase2-analysis.md
│   ├── phase3-deep-trace.md
│   └── phase4-issues.md
├── specs/
│   ├── current-implementation.md
│   └── execution-flows.md
├── solutions/
│   ├── current-solution.md
│   └── [reason]-proposed-solution.md
└── diagrams/
    └── [component]-flow.md

🏗️ Technical Guide Architect

File: skills/technical-guide-architect/SKILL.md

Creates comprehensive, agent-executable implementation guides through rigorous Socratic clarification followed by structured documentation. Always produces two deliverables: a Technical Implementation Guide and a Visual Architecture Companion.

What It Does

Transforms vague requirements into precise, bite-sized prompts an AI agent can execute sequentially — with TDD instructions, acceptance criteria, and "Do NOT" guardrails in every prompt.

Usage

Claude Code

Create a technical guide for <project description>

Examples:

Create a technical guide for building a REST API with authentication
Create an implementation guide for migrating from REST to GraphQL
Create a development playbook for a multi-tenant SaaS billing system

Cline

Using the Technical Guide Architect skill, create an implementation guide for <project>.
Begin with Phase 1 Socratic clarification — ask me the Round 1 questions before writing anything.

Important: The skill asks 12–20 clarifying questions across 4 rounds before writing any documentation. Do not skip this step — the quality of the guide depends on it.

The Clarification Rounds

Round Focus Questions
1 Project Scope End deliverable, audience, starting point, stack, timeline, compliance
2 Technical Depth Testing philosophy, security, infrastructure, deployment, integrations
3 Quality Standards Code quality, documentation, definition of done, architecture principles
4 Success Criteria What success looks like, what would make the guide insufficient

Output Deliverables

Two files are always created:

[project]-technical-guide.md     ← Executable prompts with TDD, acceptance criteria, guardrails
[project]-diagrams.md            ← Mermaid diagrams: flows, architecture, ERDs, state machines

Prompt Structure (every prompt includes)

#### Prompt N.M.P: Descriptive Title

Requirements:
- Specific, verifiable requirement

File structure:
path/
├── file1
└── file2

Write tests FIRST for:
- Test case 1

Do NOT:
- Anti-pattern to avoid

Acceptance Criteria:
- [ ] Binary pass/fail criterion

Using Both Skills Together

These skills are designed to complement each other:

Existing Codebase → [trace] → Understand current state → [technical-guide-architect] → Plan improvements
New Project       →           [technical-guide-architect] → Build from scratch

Recommended workflow for refactoring:

  1. Run @deep-trace on the area you want to change → produces ai_docs/ analysis
  2. Feed the ai_docs/solutions/ output into a technical guide request
  3. Use the generated guide to execute the refactor systematically

Example:

# Step 1
@deep-trace the payment processing module and identify coupling issues

# Step 2 (after trace completes)
Create a technical guide for refactoring the payment module based on
the analysis in ai_docs/solutions/decoupling-proposed-solution.md

Installation

Claude Code

Place the skill files anywhere in your project or global ~/.claude/ directory. Reference them using @ mentions or configure them as custom skills in your Claude Code settings.

.claude/
└── skills/
    ├── trace/
    │   └── SKILL.md
    └── technical-guide-architect/
        └── SKILL.md

Cline

Add skill content to your Cline system prompt, or reference the files directly in your task messages using Cline's file reference syntax (@file).

# In your .clinerules or system prompt:
@.claude/skills/trace/SKILL.md
@.claude/skills/technical-guide-architect/SKILL.md

Tips

  • Trace first, plan second — Never build on code you don't understand
  • Let the phases gate your work — Each phase of trace produces standalone value; stop early if needed
  • Don't skip clarification — The technical guide architect's questions prevent rework
  • Keep ai_docs/ in version control — The trace output is living documentation
  • Reuse trace output — Phase 4 issues feed directly into technical guide requirements

https://claude.ai/public/artifacts/ca74314b-91d3-482e-8329-e9105058b51a

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors