Cowork & Claude tools

At a glance

Claude Code is the standard AI development tool at Shipsy, replacing Cursor.
Cowork sessions pair Claude with a human for real-time collaboration.
Shipsy maintains custom skills and plugins (codebase-navigator, agentfleet builder, bi-skill) that extend Claude’s knowledge of Shipsy systems.
This page covers setup, the skill library, and a starter prompt kit for CS tasks.

Why this matters

Claude Code is how you’ll navigate the codebase, debug agent issues, draft evals, and explore customer configurations without needing to be a full-stack engineer. Setup takes 15 minutes; the payoff is hours saved per week.

Setup (15 min)

Prerequisites

Tool	Why you need it
GitHub account	Org access to shipsy repos
Claude Code license	Ask your manager or IT
GitHub CLI (`gh`)	Authenticates Claude Code with GitHub
Docker	Required for some local dev workflows

Steps

Go to claude.ai/code and sign in.
Connect your GitHub account.
Install Claude Code CLI: follow setup docs.
Clone the repo you want to explore: git clone https://github.com/shipsy/agent-platform.git
Open it: cd agent-platform && claude

The repo ships with a .claude/CLAUDE.md that primes Claude with Shipsy-specific context.

Key skills & plugins

Shipsy maintains custom skills that give Claude deep knowledge of Shipsy systems. Always use the right skill before asking a broad question.

Skill	What it does	When to use
`codebase-navigator`	Knows where every product area lives across all Shipsy repos	”Which repo has the routing engine?”
`agentfleet`	Guides you from a business use case to a complete CreateAgentRequest JSON	”Build me an agent for WISMO”
`bi-skill`	Knows BI datamarts, joins, field definitions, common SQL patterns	”Show me all delayed orders in Mumbai”
`slide-search`	Finds and adapts sales decks for customers	”Find a deck for a port customer”

Using a skill

> use the codebase-navigator skill to find where COD reconciliation lives

> use the agentfleet skill to build a WISMO agent for a B2C LSP

Starter prompt library for CS

Copy/paste these for common CS tasks:

Understanding a customer’s setup

> Walk me through how [agent name] is configured for [customer].
> What tools does it call, what models does it use, and where does
> it escalate?

Debugging an issue

> A customer says the agent is giving wrong tracking information.
> Where in the code does the agent fetch tracking data? Show me
> the tool call and the response parsing.

Drafting an eval

> Show me the eval format for [agent name]. Then draft 5 new
> eval scenarios covering: happy path, edge case, adversarial
> input, timeout, and language switch.

Exploring a new repo

> I'm new to this repo. Give me a 10-minute tour: where do
> agents live, where do tools live, where does the orchestrator
> live, and what's the single most important file to read first?

Building a dashboard query

> use the bi-skill. I need a dashboard showing all orders with
> 2+ failed delivery attempts in the last 7 days, grouped by
> carrier and city.

Cowork sessions

Cowork mode is a pair-programming setup where Claude and a human work on the same task in real time. Used at Shipsy for:

Complex agent feature development
Debugging cross-service issues
Onboarding new team members to unfamiliar codebases

To start a Cowork session, open Claude Code and invite a colleague via the collaboration feature.

Other useful tools

Tool	What it’s for
Granola	AI meeting notes. Records and summarizes meetings automatically.
Slack bot	Internal Slack bot for quick lookups and notifications.
Notion AI	Document search and summarization across internal docs.

Anti-patterns

Don’t paste customer data into prompts. Use anonymized samples.
Don’t run bash from a Claude session against a production DB. Local sandbox or read-only replica only.
Don’t treat Claude’s summaries as source of truth. Always verify against the actual file.

Sources

Slack: #claude-code-cowork-codex — SOP and adoption discussions
Slack: #software-factory — engineering leadership on Claude Code as standard tool
See Querying the repo with Claude Code for a guided walkthrough

Changelog

26 May 2026: Full content from Slack adoption discussions. Setup guide, skill library, and prompt kit added.

Integrations & MCP Standard CS workflows