Key highlights
1-2 hours to generate documentation for each repository
80-85% accurate on the first attempt, before SME review
One role-based HTML artifact spanning sixteen tabs
Reusable workflow across repositories in the .NET estate
Client overview
The client is a leading insurance provider that manages complex policy administration systems supporting critical business operations. With a large .NET application landscape spanning multiple interconnected repositories and limited up-to-date documentation, the organization sought to improve knowledge accessibility, accelerate developer onboarding, reduce dependency on key subject matter experts, and simplify the maintenance of business-critical applications through AI-powered code discovery and documentation.
Challenges
Reverse-engineering an undocumented 150K-line insurance platform
An insurance provider's policy administration platform spanned three deeply interconnected .NET repositories, each with 50K-70K lines of code and three to four cross-dependencies, totaling roughly 150K lines. Documentation was sparse and had fallen behind a fast-evolving codebase.
This meant that new developers took too long to become productive, while incident resolution depended heavily on a small number of people with deep platform knowledge, creating a key-person dependency. The challenge was to reconstruct this knowledge without manually piecing it together repository by repository.
Solution
Unlocking 150K Lines of Code in Hours, Not Months
Zensar built and ran a proof of concept (PoC) using Claude Code and custom Claude Skills to reverse-engineer three deeply cross-linked repositories. Claude Code reconstructed how the platform actually behaved, while Claude Skills made the workflow repeatable.
The approach generated living, role-based documentation across business context, architecture, flows, integrations, tests, and code-level detail-with outputs reusable across repositories and .NET applications. It also enabled documentation to be regenerated on demand as the codebase evolved, helping address documentation drift.
Claude Code worked directly across three cross-linked repositories, reading files and reasoning about how they connect to reconstruct how the platform actually behaved-rather than treating each repository in isolation.
Custom Claude Skills packaged the discovery and documentation workflow into reusable instructions, making the process consistent and repeatable across repositories and .NET applications without rebuilding it from scratch.
Each repository produced a single HTML artifact organized into sixteen role-based tabs, spanning executive context, architecture, journeys, flows, APIs, events, integrations, database, configuration, tests, NFRs, and codebase detail.
The approach created living documentation that could be regenerated on demand as the codebase evolved, helping address documentation drift while supporting a layered roadmap from application-level documentation to domain-level and enterprise-wide views.
Solution enablers
Tech Stack
- Claude Code
- Custom Claude Skills
- .NET
- HTML
- Swaggers
Zensar Services
- AI-accelerated code discovery
- Codebase reverse engineering
- Living technical documentation
- PoC design and execution
Business outcome
By leveraging Claude Code and reusable Claude Skills, Zensar transformed a months-long reverse-engineering and documentation effort into a repeatable workflow that delivered comprehensive, role-based documentation in hours. The approach reduced key-person dependency, accelerated developer onboarding, minimized documentation drift, and enabled documentation to be regenerated as the codebase evolved, creating a sustainable foundation for long-term maintainability.