Skip to content

Repository files navigation

Razorpay Org-Wide AI Playbook

The operating manual for Razorpay's AI builder program. A belt-progression curriculum, seven reusable Claude Code skill definitions, and a Starlight hub — all built on one Markdown source of truth.

Foundation → White → Yellow → Green → Black → Council

v0.73 alpha · updated 2026-09-10 · source of truth: Markdown
Master Index · Changelog · Roadmap · Contributing


What this is

A playbook for AI-native engineering at Razorpay. It starts before tools — before Terminal, before "what's an API" — and climbs through Foundation, four belts (White, Yellow, Green, Black), and a Staff+ Council layer for senior contributors. A reader who finishes can ship code with AI, lead a team adoption sprint, or contribute back to the org's internal skills and knowledge-base layer.

The content is Markdown-first, written to be outside-readable at the lower belts and fintech-real at the upper ones. Belts are earned by shipping, not by reading — every belt has required modules, hands-on quests, and a boss-fight capstone.

Three doors, one source

The same Markdown layer powers three consumption surfaces. Pick the one that fits how you want to read.

Door Best for How to use it
Read on GitHub Skimming, linking, and reading in the same place you work. Start at INDEX.md. Every chapter is a flat Markdown file.
Browse the hub The polished web experience with search, sidebar, and progress markers. The Astro Starlight site under hub/. See Run the hub below.
Inspect a reusable workflow Understanding how a bounded skill should behave before it is distributed. Start in the Skills Library. The seven reference definitions live under skills/; confirm the current runnable distribution in #ai-help rather than assuming Compass ships them.

Pick your starting point

  • New to software.Part 0 — Foundation. Tech 101 + Ops 101. Pre-tools, pre-AI, fully outside-readable.
  • Already know what an API, database, and deploy are.Prologue. A twelve-chapter mental-model warmup.
  • Reviewing the whole playbook as a sponsor or maintainer.Master Index.
  • Looking for templates, cards, glossary, or certification policy.Appendices.
  • Senior IC interested in the Council layer.Staff+ Council.

If you want a guided walk, use the index or hub today. The repo contains a playbook-course reference definition for a stateful Claude Code walk, but it is not currently shipped through Compass; confirm a supported runnable route in #ai-help.

Repository layout

.
├── README.md                 # you are here
├── INDEX.md                  # reader-facing master TOC
├── CHANGELOG.md              # what shipped, when
├── CONTRIBUTING.md           # design principles, content rules, voice
├── ROADMAP.md                # open questions and the v1.0 path
├── manifest.yml              # machine-readable course manifest
├── slugs.yml                 # stable URL + skill ID map
│
├── foundation/               # Part 0 — Tech 101 + Ops 101
├── prologue/                 # 12-chapter mental-model warmup
├── belts/                    # 01-white → 02-yellow → 03-green → 04-black → 05-council
├── appendices/               # A tool atlas · B setup · C skills · D issues · E roles
│                              # F slack · G glossary · H cards · I templates
│                             # J reading · L certification · N methodologies
├── case-studies/             # Boss Fight B-B case studies seed here
│
├── skills/                  # 7 Claude Code reference definitions
├── diagrams/                 # 9 signature SVG diagrams + 2 legacy aliases
├── excalidraw/               # 7 hand-drawn companion illustrations
├── hub/                      # Astro Starlight source — regenerated from above
└── scripts/                  # build + polish tooling

Run the hub locally

cd hub
npm install
npm run dev
# open http://127.0.0.1:4321/

Production build:

cd hub
npm run build       # static output → hub/dist/
npm run preview     # serve hub/dist/ locally

The hub regenerates hub/src/content/docs/ from the root Markdown on every build. Edit the root Markdown — never the generated files, which are overwritten.

Deployed hub

The hub is published to GitHub Pages on every merge to master and is accessible at:

https://razorpay.github.io/ai-playbook/

How deployments work

Trigger What happens URL
Push to master Site builds and deploys automatically (~2 min) https://razorpay.github.io/ai-playbook/
PR opened / updated Preview build deploys automatically https://razorpay.github.io/ai-playbook/pr-preview/pr-{N}/
PR closed / merged Preview is automatically removed

The bot posts a comment on every PR with the preview URL so you can verify the rendered output before merging.

Trigger a manual redeploy

Go to Actions → Deploy to GitHub PagesRun workflow → master.

Inspect the course-skill reference

The in-repo playbook-course definition specifies how a stateful Claude Code walk should pace the curriculum, read curriculum.json, and track progress in a local LEARNER.md. It is a reference implementation, not a Compass command currently available to learners.

Use the Master Index or hub for a guided reading route today. Maintainers can inspect the course-skill notes; confirm any supported runnable distribution in #ai-help before sharing invocation guidance. A future implementation may record quest and boss-fight claims, but belt awards still follow Appendix L's reviewer protocol.

What's drafted

Foundation, Prologue, all four belts, the Staff+ Council section, thirteen drafted appendices plus one planned appendix, seven Claude Code reference definitions, seven quick-reference cards, nine signature SVG diagrams, and seven hand-drawn companion illustrations. Appendix E (Roles & Forums) remains a drafted skeleton, and Appendix F (Slack Channels) is a living directory; named holders refresh quarterly. Appendix K is the root CHANGELOG.md, and Appendix M remains planned. Every other appendix is drafted end-to-end. The reading order is coherent end-to-end from first commit to Council membership.

Version history is in CHANGELOG.md. What's coming next is in ROADMAP.md.

Contribute

Read CONTRIBUTING.md before opening a PR. It covers the seven review principles, the content rules (Razorpay-term tiering, no PII, no personal names), the voice commitments, and the link-and-lint sweep that gates merges.

For substantial changes (a belt restructure, a new track, a voice change), write an AI RFC using the template in Appendix I. The RFC is reviewed by the Staff+ Council per C.3.


Next: → Master Index · → Part 0 — Foundation · → Prologue

About

No description, website, or topics provided.

Resources

Code of conduct

Contributing

Stars

5 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages