Dataplane Automation

Automation scripts and Claude AI tooling for analyzing JIRA tickets and generating implementation plans for the dataplane codebase.

🎯 Overview

This repository contains scripts that use Claude AI with Serena semantic search to:

• Fetch JIRA ticket details
• Analyze codebases semantically
• Generate implementation plans
• Build persistent knowledge over time

📁 Repository Structure

dataplane-automation/
├── scripts/               # Shell scripts for automation
│   ├── jira_plan.sh                    # Static cache approach
│   ├── jira_plan_serena.sh             # Serena non-interactive
│   ├── jira_plan_interactive.sh        # Serena interactive (recommended)
│   ├── generate-codebase-context.sh    # Generate static context
│   ├── README_SERENA.md                # Full documentation
│   └── INTERACTIVE_GUIDE.md            # Interactive workflow guide
│
├── .claude/               # Claude configuration
│   ├── skills/                         # Skill templates
│   │   ├── jira_plan.md               # Static cache skill
│   │   └── jira_plan_serena.md        # Serena skill
│   ├── agents/                         # Custom agents
│   │   └── jira-analyst.json          # JIRA analysis agent
│   └── codebase-context.md            # Cached codebase context
│
└── plans/                 # Generated implementation plans
    └── plan_*.md          # Plans organized by JIRA ticket ID

🚀 Quick Start

Prerequisites

1. Claude Code CLI installed and authenticated
2. Serena MCP plugin installed
3. Jira skill configured (/jira command available)
4. Access to the dataplane codebase at ~/dataplane

Installation

1. Clone this repo:

   git clone <repo-url> ~/dataplane-automation
   cd ~/dataplane-automation

2. Make scripts executable:

   chmod +x scripts/*.sh

3. Link scripts to your dataplane repo (optional):

   cd ~/dataplane
   ln -s ~/dataplane-automation/scripts claude-scripts
   ln -s ~/dataplane-automation/.claude .claude
   ln -s ~/dataplane-automation/plans plans

📖 Usage

Method 1: Interactive (Recommended for Complex Bugs)

cd ~/dataplane
~/dataplane-automation/scripts/jira_plan_interactive.sh ENG-825217

This opens an interactive Claude session where you can:

• Guide the investigation
• Ask follow-up questions
• Explore multiple hypotheses
• Build knowledge in real-time

See scripts/INTERACTIVE_GUIDE.md for detailed workflow.

Method 2: Automated with Serena

cd ~/dataplane
~/dataplane-automation/scripts/jira_plan_serena.sh ENG-825217

Automatically generates a plan using Serena semantic search.

Method 3: Static Cache (Fast)

cd ~/dataplane

# One-time: Generate codebase context
~/dataplane-automation/scripts/generate-codebase-context.sh

# Generate plan
~/dataplane-automation/scripts/jira_plan.sh ENG-825217

Fast analysis using pre-generated codebase context.

📊 Approach Comparison

Feature	Static Cache	Serena Automated	Serena Interactive
Speed	⚡⚡⚡	⚡⚡	⚡
Depth	⭐⭐	⭐⭐⭐	⭐⭐⭐⭐⭐
Token Cost	Low	Medium	Medium-High
Learning	❌	✅	✅✅
Exploration	❌	🟡	✅✅
Use Case	Quick/Simple	Standard bugs	Complex/Novel

🧠 Serena Features

Semantic Code Search

• find_symbol - Find classes/functions by name
• search_for_pattern - Search for patterns from logs
• get_symbols_overview - Understand file structure
• find_referencing_symbols - Track dependencies

Persistent Memory

Serena learns and remembers:

• architecture/overview - Codebase structure
• bugs/common_patterns - Known bug patterns
• Custom memories you create during analysis

📝 Generated Plans

Plans are saved to plans/plan_<JIRA-ID>.md with:

• JIRA Summary - Problem description
• Root Cause Hypothesis - Analysis with evidence
• Affected Areas - Files and line numbers
• Implementation Plan - Step-by-step fix strategy
• Test Strategy - Testing requirements

Example: plans/plan_ENG-825217.md

🔧 Configuration

Environment Variables

Scripts expect:

• ATLASSIAN_API_TOKEN - Jira API token
• ATLASSIAN_EMAIL - Your email
• ATLASSIAN_SITE - Site name (e.g., "netskope")

Claude Configuration

The .claude/ directory contains:

• Skills - Prompt templates for different approaches
• Agents - Custom agent configurations
• Context - Cached codebase information

Customization

Edit scripts to change:

• Output directory (default: plans/)
• Skill file location
• Claude CLI flags

🤝 Contributing

This is a personal automation repo, but you can:

1. Fork for your own projects
2. Adapt scripts for your codebase
3. Share improvements

📚 Documentation

• scripts/README_SERENA.md - Complete Serena documentation
• scripts/INTERACTIVE_GUIDE.md - Interactive workflow guide

🎓 Examples

Example 1: Analyze a Critical Bug

cd ~/dataplane
~/dataplane-automation/scripts/jira_plan_interactive.sh ENG-825217

# Inside the session:
You: "Search for __skip_object_scan in the logs"
You: "Find where this flag is set"
You: "Show me the inspect_cibuffer function"
You: "Save this finding to memory"
You: "Generate the implementation plan"

Example 2: Batch Process Tickets

for ticket in ENG-12345 ENG-12346 ENG-12347; do
  ~/dataplane-automation/scripts/jira_plan_serena.sh $ticket
done

Example 3: Continue Previous Analysis

cd ~/dataplane
claude --continue --agent jira-analyst

# Then:
"Continue analyzing ENG-825217 with the new information"

🔍 Troubleshooting

"No active project" Error

Make sure you're in the dataplane directory when running scripts.

Scripts Not Executable

chmod +x ~/dataplane-automation/scripts/*.sh

JIRA Command Not Found

Verify the eng-skills plugin is installed and /jira works.

Serena Errors

Activate the project:

cd ~/dataplane
claude
# Then: "Activate the dataplane project with Serena"

📄 License

Personal automation scripts. Use at your own discretion.

🙏 Credits

• Claude AI by Anthropic
• Serena MCP for semantic code search
• eng-skills for JIRA integration

---

Happy Bug Hunting! 🐛🔍