Implementation Tracker Agent for Claude Code

A context-aware implementation tracker agent for DevOps, Platform Engineering, and SRE tasks. This agent automatically discovers tools in your codebase, learns from your previous patterns, and generates detailed, phase-based implementation trackers.

Features

✅ Dynamic Tool Discovery: Automatically detects 50+ DevOps tools (Terraform, Kubernetes, Docker, CI/CD, monitoring, etc.)

✅ Context-Aware: Scans conversation, codebase, docs, git history, and previous trackers

✅ Pattern Learning: Adapts to your team's established patterns and style

✅ Interactive: Uses questions to validate understanding and gather preferences

✅ Detailed Generation: Creates specific tasks with real commands, paths, and validation checks

✅ Self-Updating: Maintains and updates a tools catalog after each run

What Gets Installed

1. Agent File (~/.claude/agents/implementation-tracker.md)

   • The main agent that powers tracker generation
   • Shows up in @agent autocomplete

2. Template File (implementation-tracker-prompt.md)

   • Defines the structure and rules for implementation trackers
   • Includes guardrails, phase gates, and validation requirements

3. DevOps Tools Catalog (~/.claude/data/devops-tools-catalog.json)

   • Tracks discovered DevOps tools across your repos
   • Stores learned patterns from your previous trackers
   • Updates automatically with each agent run

4. Working Directory (configurable, default: ~/working-docs/)

   • Where your implementation trackers are saved
   • Local-only directory (not git-managed)

Installation

Prerequisites

• Claude Code installed and configured
• Bash shell (macOS/Linux)

Quick Install

1. Download/clone this repository
2. Run the installation script:

cd implementation-tracker-setup
chmod +x install.sh
./install.sh

3. Follow the prompts to configure your working-docs directory

Manual Install

If you prefer to install manually:

1. Copy files to the correct locations:

# Create directories
mkdir -p ~/.claude/agents
mkdir -p ~/.claude/data
mkdir -p ~/working-docs  # or your preferred location

# Copy agent
cp implementation-tracker.md ~/.claude/agents/

# Copy template (adjust path as needed)
cp implementation-tracker-prompt.md ~/implementation-tracker-prompt.md

# Copy catalog
cp devops-tools-catalog.json ~/.claude/data/

2. Edit ~/.claude/agents/implementation-tracker.md:

   • Replace TEMPLATE_PATH_PLACEHOLDER with your template path
   • Replace WORKING_DOCS_PATH_PLACEHOLDER with your working-docs path

3. Verify installation by checking these files exist:

   • ~/.claude/agents/implementation-tracker.md
   • ~/.claude/data/devops-tools-catalog.json
   • Your template and working-docs directories

Usage

Invoking the Agent

1. Open Claude Code
2. Type: @agent
3. Select: agent-implementation-tracker
4. Describe your project/task

Example Session

You: @agent implementation-tracker

You: I need to migrate our PostgreSQL database from RDS to Aurora

Agent:
[Scans repos for database configs, Terraform, K8s manifests]
[Reads previous trackers to learn your patterns]

Agent: "Based on context analysis, I detected:

Project Type: Database Migration (PostgreSQL → Aurora)
Affected Environments: staging, production
Technologies Found: Terraform, Kubernetes, AWS RDS, Helm

Is this understanding correct?"

[You confirm]

Agent: "What guardrails should I apply?"
[Shows options for approval gates, update frequency, evidence]

[You select preferences]

Agent: "Proposed phases:
- Phase 0: Plan & Decision
- Phase 1: Aurora Cluster Setup (Staging)
- Phase 2: Data Migration Testing
- Phase 3: Production Migration
- Phase 4: Validation & Rollback Testing
- Phase 5: Documentation & Knowledge Transfer

Does this make sense?"

[You approve or modify]

Agent: "Where should I save this tracker?"
[Default: ~/working-docs/ or custom path]

Agent: Generates detailed tracker with:
✓ Actual commands from your configs
✓ Real database names from your K8s secrets
✓ Specific validation queries
✓ Rollback procedures

"Tracker saved: ~/working-docs/postgresql-aurora-migration-implementation-tracker.md"

How It Works

Phase 1: Context Discovery

• Conversation Analysis: Extracts objectives, technologies, constraints from chat
• Codebase Scanning: Discovers IaC (Terraform, K8s, Helm), CI/CD, containers, monitoring tools
• Documentation Mining: Reads README files, architecture docs, previous trackers
• Git Analysis: Reviews recent commits, branches, change patterns
• Catalog Loading: Reads previously discovered tools and patterns

Phase 2: Interactive Confirmation

Uses questions to validate:

• Project understanding
• Guardrail preferences
• Phase structure
• Save location
• Any ambiguities

Phase 3: Tracker Generation

• Applies template structure
• Generates specific, actionable tasks with real commands
• Includes actual file paths and tool names discovered
• Adapts based on learned patterns from your previous work

Phase 4: Catalog Update

• Updates tools catalog with newly discovered tools
• Records learned patterns for future runs
• Tracks statistics on tracker generation

What Tools Are Discovered?

The agent automatically detects 50+ DevOps tools across categories:

Infrastructure as Code

• Terraform, Kubernetes, Helm, Ansible, CloudFormation, Pulumi

Containers & Orchestration

• Docker, Podman, Docker Swarm

CI/CD

• GitHub Actions, GitLab CI, Jenkins, ArgoCD, CircleCI, Tekton

Cloud Providers

• AWS, GCP, Azure configs

Monitoring & Observability

• Prometheus, Grafana, Datadog, New Relic, ELK Stack, Jaeger

Service Mesh

• Istio, Linkerd, Consul

Configuration Management

• Puppet, Chef, SaltStack

Secrets Management

• Vault, Sealed Secrets, SOPS

Networking

• Nginx, Traefik, HAProxy

Databases

• PostgreSQL, MySQL, MongoDB, Redis

Message Queues

• Kafka, RabbitMQ, NATS

And more! The agent dynamically discovers any tool configuration files in your repos.

Tracker Structure

Generated trackers include:

1. Title & Overview

   • Objectives, success criteria, scope
   • Progress tracking (auto-calculated from checkboxes)

2. Guardrails (Process Gates)

   • Phase gating rules
   • Blocker handling procedures
   • Update requirements

3. Status Overview

   • List of all phases with status

4. Decision Record

   • Options considered, chosen path, rationale
   • Risks, mitigations, rollback approach

5. Phases (customized per project)

   • Each phase has:
     • Objective
     • Time estimate
     • Prerequisites
     • Checklist with concrete tasks
     • Expected outcomes
     • Phase gate

6. Issue/Blocker Log

   • Structured cards for tracking blockers
   • Resolution plans and evidence

7. Appendices

   • Useful commands, snippets, change log

Tips for Best Results

1. Provide Context

The more context you provide, the better:

• Mention specific technologies you're using
• Reference relevant docs with @file.md
• Describe environments (dev, staging, prod)
• State any constraints or deadlines

2. Review & Customize

Generated trackers are starting points:

• Review the proposed phases
• Adjust time estimates based on your knowledge
• Add team-specific requirements
• Customize validation checks

3. Keep Trackers Updated

Follow the guardrails:

• Update after each task completion
• Check off completed tasks
• Fill in Notes/Evidence with outputs
• Update progress percentage
• Log blockers immediately when they occur

4. Learn & Improve

The agent learns from your previous trackers:

• Keep old trackers in working-docs/
• Use consistent naming and patterns
• The more you use it, the better it adapts

Troubleshooting

Agent doesn't appear in @agent list

• Verify file exists: ls -la ~/.claude/agents/implementation-tracker.md
• Check frontmatter format (YAML with --- delimiters)
• Restart Claude Code

Agent can't find template

• Check template path in agent file
• Verify template exists: ls -la ~/implementation-tracker-prompt.md
• Update path in ~/.claude/agents/implementation-tracker.md

No tools discovered

• Ensure you're in a directory with infrastructure code
• Check file patterns match your setup
• Review catalog: cat ~/.claude/data/devops-tools-catalog.json

Generated tracker is too generic

• Provide more context in your request
• Reference specific files with @filename
• Let agent scan your codebase first
• Run more sessions to help agent learn your patterns

Customization

Change Default Working-Docs Location

Edit ~/.claude/agents/implementation-tracker.md and find the line:

"Default: /path/to/working-docs/"

Update to your preferred path.

Modify Template Structure

Edit your implementation-tracker-prompt.md file to customize:

• Phase naming conventions
• Section structure
• Guardrail rules
• Required evidence formats

The agent will use your updated template automatically.

Add Custom Tool Patterns

Edit ~/.claude/agents/implementation-tracker.md and add patterns to the tool discovery section:

**Your Custom Tool:**
- CustomTool: `**/custom-config.yaml`, `**/custom/*.conf`

File Locations Reference

~/.claude/
├── agents/
│   └── implementation-tracker.md        # The agent
└── data/
    └── devops-tools-catalog.json        # Tools catalog

~/                                       # Or your chosen location
├── implementation-tracker-prompt.md    # Template
└── working-docs/                       # Generated trackers
    ├── project1-implementation-tracker.md
    └── project2-implementation-tracker.md

Support & Feedback

This agent was designed for DevOps, Platform Engineering, and SRE workflows. If you encounter issues or have suggestions:

1. Review this README for troubleshooting tips
2. Check the generated tracker structure matches your needs
3. Customize the template to fit your team's style
4. Review the agent file for tool discovery patterns

License

Free to use and modify for your needs.

---

Happy implementing! 🚀