Skip to content

Appendix F: Slack Channels & Rules of Engagement

Status: drafted · Time: 6 min · Audience: everyone Outcome: Find the right channel for the question you have, follow the etiquette so the program’s primary channels stay useful, and know where to celebrate wins or escalate concerns.

This is the live directory. Channel handles, purposes, and owners as of the date in the frontmatter. The directory is reviewed quarterly and on-demand when channels split or merge. If a row looks stale, ping #ai-help and it gets fixed in the next revision.

The directory of Slack channels the playbook references, paired with the etiquette and rules of engagement that keep them useful. Channels are organised by purpose, not alphabetically. Every chapter that says “post in [channel]” links here.

The first channels every builder needs.

The former #claude-onboarding-support channel was archived on 2026-05-16 with instructions to log further support in #ai-help; the rows below use the active handle.

HandlePurposeOwner / first respondersFirst-response expectation
#ai-helpMyAccess access requests, install/setup help, LiteLLM gateway issues, manager-OOO approval bypass, troubleshooting claude errors, and general AI help. The active successor to archived #claude-onboarding-support for White Belt builders.AI help responders; access escalations to @techit subteam< 1 day business
#engineering-allOrg-wide engineering announcements. Home of the canonical Claude rollout thread. Don’t post setup questions here — route to #ai-help instead.Engineering leadershipAnnouncements, not Q&A
#devstack-onboarding-supportRazorpay devstack onboarding — local environment, repo bootstrap, devstack shrc. Adjacent to Claude setup; sometimes the right place when the error is environment-level rather than Claude-level.Devstack team< 4 hours business

Where builders gather to swap patterns, share wins, and showcase what they’ve built. The default rooms for “look what I made” conversation.

HandlePurposeOwnerFirst-response expectation
#ai-bulletinMagical AI stories — share what you’re doing with AI; learn from each other. The org-wide signal channel for AI wins. Use threads for follow-up questions.Khilan HariaAsync
#ai-code-championsHacks, ideas, and best practices for getting the most out of AI coding tools. The peer-to-peer craft room for builders shipping with AI daily.Yash DoshiAsync
#ai-coding-infraAI coding infrastructure — model access and routing, usage and cost, and shared AI coding tooling.Kaushik BhatAsync
#ai-pmmAI for PMM and marketing — campaign workflows, content generation, audience analysis. The PM-Marketing-adjacent home.Vaidehee Prayaag JoshiAsync
#product-ai-labsProduct AI Labs pilot — Claude Code + Compass first cohort. Show-and-tell and knowledge sharing for the product team.Product Spec Reviewer; Aravinth P K for PM-side coordinationAsync
#pod-velocity-and-ai-working-groupThe cross-pod working group on velocity + AI. Coordination, decisions, working-group artefacts (tracker in pinned).Nirvisha MankadPer-cadence
#rzp-claude-skillsRazorpay Claude Skills — share skills you’ve built, find skills you can reuse. The discovery surface for the Skills Library.Shobhit JainAsync
#devex-skillsExperiments toward Software AGI with Agent Skills. Showcase ideas and experiments — scan /agent-skills and /claude-plugins repos before building, in case it already exists.Vikas NaiduAsync
#cowork-helpCowork desktop app — setup, plugin install, connector pairing, troubleshooting. Adjacent to but distinct from Claude Code.Anurag Rastogi< 1 day business

For builders on the design lane (designers, design engineers, anyone doing Figma → code work).

HandlePurposeOwnerFirst-response expectation
#design-systemBlade — the canonical design-system channel. Component questions, contribution patterns, the “Everything related to Blade!” channel.Saurabh Soni and the design-system leads< 1 day business
#experience_fe_coreFrontend Universe + Blade support. Raise a support ticket from the channel bookmarks for routed help.Varun Achar and the frontend-core teamPer-ticket SLA
#blade-consumerBlade for consumer apps — narrower-scope Blade questions specific to consumer surfaces.Ravikumar R< 1 day business
#ai-in-designAI in design practice — how designers are using AI tools in their day-to-day, prompts that work, gotchas.Saurabh SoniAsync
#ai-squad-designThe AI design squad — the focused group driving AI design work.Abhinav Krishna AadirajuAsync

For builders contributing back to platform layers — MCP servers, API design, the devex platform itself.

HandlePurposeOwnerFirst-response expectation
#developer-experienceInnovating on how Razorpay writes code, tests, and automates reviews. The home channel for devex-platform work that the playbook depends on.Kaushik Bhat< 1 day business
#mcp-devRazorpay’s internal MCP server development. Links to github.com/razorpay/razorpay-mcp-server. The home for building and consuming Razorpay-specific MCP connectors.Shalky Sharma< 1 day business
#testing-slashTesting the /slash plugin patterns — slash-command development, regression checks, plugin POCs.Arnav GaurAsync
#ai-pr-merge-requestsWhere AI-assisted PRs surface for visibility and review-routing. Use when you’ve shipped AI-generated work and want it found by reviewers tracking the pattern.Vaibhav DhirAsync
#api_councilThe API Council — submissions, reviews, design-guide questions. Referenced in B.15 — API Council contributions.Khilan HariaPer-review cadence

For PMs and analysts using Analytics Agent or validating certified metrics.

HandlePurposeOwnerFirst-response expectation
#analytics-self-serveAnalytics Agent onboarding, certified-metric questions, redesign shadow comparisons, sign-off, and rollout escalations. Keep each domain’s evidence in its existing thread when one is linked.Analytics Self Serve teamPer-validation cadence

Channels every product person should be in even outside the playbook program. Not playbook-specific, but the discoverability surface for what’s happening across the org.

HandlePurposeOwnerFirst-response expectation
#product-allOrg-wide PM channel — cross-pod product conversation, PM-of-the-week patterns, product-leadership signal.Chirag JainAnnouncements + Q&A
#product-function-announcementsProduct-function announcements and enablement — PM workshops, operating updates, and rollout asks.Product-function announcement adminsAnnouncements only
#prod-bulletinCustomer-facing product announcements only. Screenshots and demo links required; no discussions. The signal channel for “what just shipped externally.”Shashank MehtaAnnouncements only
#product-design-bulletinProduct design bulletin — design launches, pattern updates, design-system shipped work. Follow the guidelines.Saurabh SoniAnnouncements only

For builders evaluating tools outside the canonical stack.

HandlePurposeOwnerFirst-response expectation
#vibe-coding-tool-pocsPOCs with external tools — Superblocks, Replit, Loveable, etc. Goal: find a tool that works with Razorpay’s design system, frontend, backend, and deployment stack.Chaitanya Vikas DeorukhkarAsync

A new channel is justified when:

  • A recurring topic does not fit any existing channel.
  • A specific surface (a plugin, an integration, a forum) needs its own coordinated home.
  • A cohort or initiative has enough sustained traffic to outgrow the cohort or program-flagship channel.

The flow:

  1. Post in #ai-help describing the proposed channel and its purpose.
  2. The program lead (or delegate) approves and creates the channel.
  3. Send a one-line PR adding the channel to this appendix; the directory updates on the next merge.

The program’s primary channels stay useful when the etiquette is followed. Six rules.

1. Ask publicly, not in DM. A question asked in the right public channel becomes searchable for the next person who hits the same problem. A question asked in DM helps you and dies. Resist the DM impulse; post publicly. The cost to you is brief; the benefit to the next person is real.

2. Post the full output, not “it’s broken”. When something fails, the diagnostic information is what enables a quick answer. Post the command you ran, the full error message, the relevant config (with secrets redacted), and what you have already tried. “Claude is failing” is unhelpful; “I ran <command> and got <exact error>; here is the redacted config and what I tried” is helpful.

3. Celebrate merged PRs with the link. The celebrations channel works when wins are visible. Post the PR link, name what shipped, and tag the team if relevant. A celebrations channel of vague humble-brags is worse than no channel; the link is the discipline.

4. Never paste regulator-protected data. Per G.22 and G.24. Customer data, payment card information, secrets, and tokens never go in any channel — public or private. If you need to share a real-world example, redact aggressively or use a synthetic equivalent.

5. When you solve a problem, contribute to Appendix D. The fix that worked for you helps the next person only if it lands in Appendix D — Known Issues + FAQ. The contribution flow is documented there. A fix that lives only in a Slack thread evaporates within months.

6. Tag the right channel; do not cross-post. A question that fits #design-system goes in #design-system. Cross-posting (posting the same question in three channels) signals urgency to no one in particular and dilutes attention. If you genuinely do not know which channel is right, ask in #ai-help and let it route.

Quick reference for “where does this question go?”. Full role-and-forum mapping in Appendix E.

SymptomFirst contact
Can’t install Claude Code; claude --version not found; setup script erroring#ai-help
Can’t see Claude AI in MyAccess; manager OOO blocking approval#ai-help with @techit tagged
Have Claude access but no LiteLLM key, no enabled models, or key can only access models=[...]#ai-help — ask for LiteLLM gateway provisioning or model enablement; see D.13
403 PERMISSION_DENIED referencing aiplatform.googleapis.com; stale Vertex env vars#ai-help — see W.5 failure mode #3
An enabled Opus route returns exceeded budget, or the total LiteLLM budget is exhaustedA named enabled route with exceeded budget has hit its model cap; check usage and use a lower-cost enabled route for routine work. Do not infer that Opus 4.6/4.7 retired or switch to 4.8 solely from that message. If the total budget is exhausted, wait for reset or post an approved business blocker in #ai-help with manager approval visible; see D.5 and D.6.
Razorpay devstack / local environment broken#devstack-onboarding-support
Cowork desktop app issue#cowork-help
Blade component question; design-system compliance#design-system or #experience_fe_core ticket
MCP connector failing or missing for a Razorpay service#mcp-dev
Analytics Agent onboarding, certified-metric comparison, or redesign sign-off#analytics-self-serve
API design question; API council submission#api_council
Security concern; sensitive data in a prompt; redline-flagged reviewSecurity review owner directly — see Appendix E
”Show me what you built”; share a skill#product-ai-labs or #rzp-claude-skills

When new channels are created, the naming follows a small pattern.

  • Prefix conventions: program- for program-wide channels, cohort- for cohort channels, the team or surface name for surface-specific channels.
  • Lowercase, hyphenated.
  • Names describe what is in the channel, not who is in it.
  • Avoid duplicating purpose: if a similar channel exists, ask whether the new channel is needed.

The actual prefix conventions are owned by the program’s communication team and may evolve. The principles above are durable.

Channels change. Handles rename, ownership rotates, new initiatives spawn new rooms. The discipline that keeps this page useful:

  • Quarterly review. The full directory is re-walked once per quarter. Each row is checked: is the channel still active, is the owner still the owner, is the purpose still accurate. The frontmatter updated date is refreshed.
  • On-demand patches. When a channel splits, merges, archives, or rotates owner, anyone can open a one-line PR against this appendix. Ping #ai-help for the lazy path.
  • No silent staleness. If a row is wrong, it is the appendix’s problem, not the reader’s. Staleness is a bug; report it like any other.

The durable pieces of this page — the etiquette, the escalation pattern, the request-a-channel flow — change once a year at most. The volatile pieces — handles, owners, first-response expectations — are the ones the discipline above is built to keep honest.


Last reviewed: 2026-09-12. Revision cadence: quarterly, plus on-demand. If a row is stale, ping #ai-help and this page gets patched.