Claude Kit
A Claude Code starter kit built for Azure DevOps teams — agents, hooks, MCP servers, and slash commands that automate the full work item → branch → PR → release → deploy lifecycle. Installed into any project with a single
npxcommand.
About
Claude Kit installs a complete AI development infrastructure into any project that uses Azure DevOps for work tracking, source control (Azure Repos), and CI/CD (Azure Pipelines). It includes specialized agents that handle specific tasks (deployment, database ops, code review), security hooks that prevent mistakes automatically, MCP server connections to your tools (Azure DevOps, databases, Teams, Stripe, Azure CLI), and slash commands that automate the entire workflow from work item to pull request.
The slash commands assume Azure DevOps as the system of record:
- Work items — read, created, and updated via the Azure DevOps MCP server (
AB#1234ids throughout) - Branches & pull requests — created in Azure Repos and voted on / merged through Azure DevOps
- Releases & deployments — tracked as Azure DevOps iterations (
Release #N) and tags; CD runs on Azure Pipelines - Wiki, test plans, advanced security alerts — all surfaced through the same Azure DevOps MCP server
Not using Azure DevOps? The Claude Code primitives (agents, hooks, memory) are still useful, but the slash commands and the deployment workflow won't apply out of the box — you'd need to rewrite the
/implement,/review,/deploy,/create-release,/deploy-release,/cherry-pick,/promote,/rollback,/track,/rework, and/resolve-feedbackcommands against GitHub / GitLab / Jira / etc.
Every session Claude learns from your feedback and gets better at helping you specifically. The infrastructure is modular — install only what your project needs.
What's Included
Prerequisites
- Claude Code CLI installed
- Node.js 18 or later
- An Azure DevOps organization with a project for work items, an Azure Repos Git repository, and (optionally) Azure Pipelines for CD
- An Azure DevOps Personal Access Token (PAT) — used by the Azure DevOps MCP server. Scopes needed: Work Items (read/write), Code (read/write), Build (read/execute), Wiki (read/write)
Verify it works
npx @chris1807/claude-kit --help
You should see the help output. The package is published publicly on npmjs.com — no authentication or .npmrc setup required.
Installation
How It Works
Every install mode does two things:
- Global agents → installed to
~/.claude/agents/(works across ALL your projects) - Project files → installed to the target project directory (agents, hooks, commands, MCP config, settings)
Option 1: Interactive Install (Recommended)
Best for first-time setup or when you want to pick exactly what you need.
npx @chris1807/claude-kit init
You'll be asked:
- Target directory — where is your project?
- Components — checkboxes to pick which parts to install:
- ☑ Project Agents (deployer, db-admin)
- ☑ Hooks (secret blocker, auto-format, etc.)
- ☑ Slash Commands (/implement, /review, /resolve-feedback, /deploy, /create-release, /deploy-release, /add-to-release, /cherry-pick, /promote, /rollback, /track, /cleanup-branches, /quote, /explain)
- ☑ MCP Servers
- ☑ Settings
- ☑ CLAUDE.md Workflow
- ☑ .gitignore Updates
- Database type — MongoDB, SQL Server, Azure SQL, PostgreSQL, or None
- MCP servers — pick which ones:
- ☑ Playwright (browser testing)
- ☑ Microsoft Teams (notifications)
- ☐ Stripe (payments) — off by default
- ☑ Azure CLI (infrastructure)
- Existing files — for each file that already exists, choose overwrite or skip
Option 2: Install to a Specific Project
Same as Option 1, but you specify the project path upfront:
npx @chris1807/claude-kit init /path/to/your/project
Option 3: Install Everything (Minimal Prompts)
Installs all components. Only asks which database type (there's no sensible default).
npx @chris1807/claude-kit init --all
This installs:
- ✅ Global agents (backend, frontend, legacy, mockup, reviewer, test-runner, build-validator, lint-checker, azure-ops, security-auditor)
- ✅ Project agents (deployer, db-admin)
- ✅ All 10 hooks
- ✅ All 13 slash commands
- ✅ MCP servers: Playwright, Teams, Azure CLI (+ your DB choice)
- ✅ Settings, CLAUDE.md workflow, .gitignore
- ❌ Stripe (not included in --all, add via interactive mode)
- ⏭️ Skips files that are already identical
- 🔄 Overwrites files that have changed
Option 4: Fully Automated (Zero Prompts)
Specify the database type as a flag — no prompts at all:
npx @chris1807/claude-kit init --all --db=mongo
npx @chris1807/claude-kit init --all --db=mssql
npx @chris1807/claude-kit init --all --db=azuresql
npx @chris1807/claude-kit init --all --db=postgres
You can also target a specific directory:
npx @chris1807/claude-kit init /path/to/project --all --db=mongo
Option 5: Global Agents Only
Just installs the 13 global agents to ~/.claude/agents/. No project files, no prompts.
npx @chris1807/claude-kit init --global-only
Use this when you just want the global agents on a new machine and will install project files separately per project.
Re-Running the Installer
Safe to run as many times as you want. The installer:
- Skips files that are identical (no unnecessary changes)
- Asks before overwriting files that have changed (in interactive mode)
- Overwrites changed files silently (in
--allmode) - Merges MCP server configs (adds missing servers without removing existing ones)
Installing on Additional Projects
Already installed on one project? Just run the same command for the next one:
npx @chris1807/claude-kit init /path/to/another/project
The installer automatically skips global agents that are already installed (they're identical) and only installs the project-specific files. You don't need --global-only or any special flag — it just works.
📦 Global Agents → ~/.claude/agents/
= azure-ops.md (identical, skipped) ← already installed, skipped
= security-auditor.md (identical, skipped)
= reviewer.md (identical, skipped)
📦 Project Agents → .claude/agents/ ← fresh install for this project
✓ deployer.md
✓ db-admin.md
...
What Gets Installed
~/.claude/agents/ ← Global (all projects)
├── backend.md # .NET/C# backend developer (Clean Architecture)
├── frontend.md # React/TypeScript frontend developer
├── legacy.md # Lucee/CFML legacy app developer (RBWO + others)
├── mockup.md # HTML mockup designer
├── reviewer.md # Code reviewer (read-only)
├── test-runner.md # Test executor — xUnit, Vitest, Playwright (read-only)
├── qa.md # Browser QA + bug repro on a running environment (read-only)
├── build-validator.md # Build checker (read-only)
├── lint-checker.md # ESLint + dotnet format checker
├── azure-ops.md # Azure infrastructure management
└── security-auditor.md # Security scanning (read-only)
your-project/ ← Project-specific
├── .claude/
│ ├── agents/
│ │ ├── deployer.md # Commit → push → deploy → monitor
│ │ └── db-admin.md # Database queries and data management
│ │
│ ├── hooks/
│ │ ├── sensitive-data-blocker.sh # BLOCKS mongosh queries for TIN/SSN/PII
│ │ ├── sensitive-data-mcp-blocker.sh # BLOCKS MCP DB queries for TIN/SSN/PII
│ │ ├── sensitive-data-output-blocker.sh # BLOCKS output containing PII fields
│ │ ├── secret-blocker.sh # BLOCKS hardcoded secrets before write
│ │ ├── protected-files.sh # BLOCKS/warns on critical file edits
│ │ ├── auto-format.sh # Auto-runs formatters after edits
│ │ ├── test-on-change.sh # Suggests related tests after edits
│ │ ├── work-item-intent.sh # Routes "create a bug/story" prompts to /create-work-item
│ │ ├── uat-reminder.sh # Reminds to run UAT after features
│ │ └── self-improve.sh # Saves learnings after each session
│ │
│ ├── commands/
│ │ ├── implement.md # /implement AB#1234
│ │ ├── autopilot.md # /autopilot AB#1234
│ │ ├── review.md # /review 142
│ │ ├── deploy.md # /deploy "commit message"
│ │ ├── create-release.md # /create-release 23
│ │ ├── deploy-release.md # /deploy-release 23 staging
│ │ ├── cherry-pick.md # /cherry-pick AB#1234 prod
│ │ ├── promote.md # /promote main dev
│ │ ├── rollback.md # /rollback AB#1234 prod
│ │ ├── add-to-release.md # /add-to-release 24 AB#4599
│ │ ├── track.md # /track release 24
│ │ ├── cleanup-branches.md # /cleanup-branches
│ │ ├── close-orphan-tasks.md # /close-orphan-tasks
│ │ ├── quote.md # /quote AB#1234
│ │ └── explain.md # /explain AB#1234
│ │
│ └── settings.json # Hook and MCP registration
│
├── .mcp.json # MCP server configuration
└── CLAUDE.md # Gets workflow section appended
Usage
Start Working
cd /path/to/your/project
claude
Verify MCP servers are connected:
/mcp
Ultracode (multi-agent workflows)
The kit's agents normally run as sequential delegation — the main loop hands work to one specialist at a time through the Task tool. Ultracode is a separate, harness-level gear that authorizes the Workflow tool to fan out many agents in parallel under a deterministic script. The two compose: ultracode doesn't replace the pipeline, it parallelizes the parts that are embarrassingly parallel.
Ultracode is opt-in. It's on only when you type ultracode in a prompt, when a system-reminder confirms it, or when a slash command's instructions say to use Workflow. When it's off, everything runs the normal sequential way — nothing changes.
Only the main Claude Code loop and slash commands can call Workflow. Subagents (anything running under Task) cannot — so ultracode-scale fan-out is a main-loop / slash-command concern.
Kit operations that benefit from ultracode when it's on:
Short, single-query operations (/track, /explain, /close-orphan-tasks) don't need ultracode — they're already one pass and gain nothing from fan-out. Reach for it when the work-list is large and the per-item work is independent.
Implement a Work Item
/implement AB#1234
Claude automatically:
- Reads the work item from Azure DevOps
- Summarizes and asks you to confirm — shows title, description, acceptance criteria. Waits for you to agree or add context
- Creates a feature branch based on work item type
- Explores the codebase and presents an implementation plan — lists files to create, modify, and delete. Waits for your approval before writing any code
- Implements using backend/frontend agents (only after plan approval)
- Runs tests, linting, and build validation
- Generates a UAT checklist from acceptance criteria
- Pauses for you to manually test
- Creates the PR after you confirm
- Attaches any hand-run deployment script to the work item, with the run command in the attachment comment, so
/promote,/cherry-pick, and/deploy-releaselist it for every environment. Data fixes are planned as run-once migrations the app applies by itself wherever the project has that mechanism; a hand-run script is the fallback and the plan says why
Implementing a whole Feature: point /implement at a Feature work item and it implements the child User Stories in waves driven by the Custom.Order field — stories with the same order value are implemented in parallel (one agent per story, each in an isolated git worktree), and the next order value starts only after the previous wave is merged and green. Everything lands on one feature/ branch with a single review, UAT pass, and PR that links every story.
Implement Without Stopping
/autopilot AB#1234
Runs every step of /implement — plan, architect review, branch, Task, implementation, build, tests, lint, code review, PR — but never stops to ask. Where /implement would wait for you, Claude makes the call and logs it. An independent architect-review agent stands in for your plan approval, and the verified find → refute review stands in for your fix decisions. The PR is where you come in: it's labeled autopilot, its first comment thread holds the plan and every decision, and its description carries the UAT checklist nobody has run yet. Autopilot never merges, deploys, or moves an item past Code Review.
It only stops before writing code (item not found, already has a PR, nothing to build, worktree locked by another session). If it can't get the build, tests, or review clean in three rounds, it opens a draft PR with what's left and leaves the item Active. Set /effort before starting, and run in auto mode (or /add-dir the worktree) so Claude Code's own permission prompts don't stall it.
Review a Pull Request
/review 142
Claude automatically:
- Reads the full PR diff
- Reads the linked work item and checks all acceptance criteria
- Reviews for Clean Architecture, security, missing tests, code quality
- Posts inline comments on all findings
- Posts a PR-level summary
- Asks: "Approve, Request Changes, or skip the vote?"
Deploy Changes
/deploy "Add payment export feature"
Claude automatically:
- Runs pre-flight checks (dotnet build, tsc)
- Stages and commits with the provided message
- Pushes the current branch
- Triggers the CD pipeline if on an environment branch (
dev,test,staging,prod) — never onmain, which has no pipeline
Create a Release
/create-release 23
Claude automatically:
- Asks which work items to include
- Creates a
Release #23iteration in Azure DevOps - Assigns and tags all work items with
release-23
Deploy a Release
/deploy-release 23 staging
Claude automatically:
- Finds all work items in Release #23 and checks each against the environment's gate state (
Ready for Staginghere) - Cherry-picks their commits into
release/23-to-staging - Creates a PR targeting the
stagingbranch — merging it triggers the Staging pipeline - Links all work items to the PR
- Asks who verifies each product group on Staging (
COM,PAY, …) — candidates are the non-developer names already on those work items - Waits for Serena (the AI reviewer) to approve the PR, then completes it — for
dev,test, andstagingonly; aprodPR you merge yourself - Watches the Staging pipeline; once it's green, moves the work items to
Stagingand assigns them to the approved verifiers
The same command walks a release through test → staging → prod; with no environment given it picks the next one.
Cherry-Pick Work Items
/cherry-pick AB#1234 AB#1235 prod
Cherry-picks specific work items to an environment without a formal release. Same gate check, same verifier question, and the same automatic state advance + assignment once the pipeline is green as /deploy-release. Targeting dev, test, or staging it also merges the PR itself once Serena approves it; a prod PR waits for a person. main is also a valid target — that's how a hot fix that went straight to prod is brought back into the compare branch.
Promote an Environment
/promote main dev
/promote staging prod
/promote ← auto-detects from the current branch
Creates a PR to promote all code from one branch to the next in the chain (main → dev → test → staging → prod). Shows every commit and work item before confirming, gate-checks the work items, asks who verifies each product group, and advances and assigns them once the merge's pipeline is green. Below production it also merges the PR for you: it polls until Serena (the AI reviewer) approves, then completes it with autocomplete — a PR into prod is always merged by a person. main → dev is the usual first hop after feature PRs merge — nothing has deployed before that.
Rollback a Deployment
/rollback AB#1234 prod
/rollback last staging
Reverts specific commits or the last deployment on an environment. Creates a revert branch and PR.
Add Work Items to a Release
/add-to-release 24 AB#4599 AB#4600
Adds work items to an existing release — assigns them to the iteration and tags them.
Check Status
/track release 24
/track pipeline
/track AB#4521
/track staging
/track
Shows the status of a release, pipeline, work item, environment, or a high-level overview of everything.
Quote a Work Item
/quote AB#1234
Displays a work item as a formatted blockquote — title, type, state, assignee, description, acceptance criteria, and child items. Useful for pasting into PR descriptions, Teams messages, or discussions.
Explain a Work Item
/explain AB#1234
Reads a work item and explains it in plain language:
- Summary — what the work item is about in 1–2 sentences
- What needs to happen — acceptance criteria translated into concrete actions
- Why it matters — business/user value
- Current status — state, assignee, parent, child progress, linked PRs
- Scope & risks — flags ambiguity, missing criteria, or large scope
Clean Up Merged Branches
/cleanup-branches
/cleanup-branches --dry-run
Finds and deletes branches that have been fully merged. Protects environment branches. Use --dry-run to preview without deleting.
Branching Strategy
Target State (All Projects)
main is the compare branch — the default branch, the one feature PRs target, and the one every environment is compared against. Merging into main deploys nothing. Each environment has its own branch, and a PR merging into that branch is what triggers its CD pipeline:
Note: A project that hasn't migrated yet (e.g. one that still has only
mainanddevelop) keeps working — every command reads the project's Pipeline Configuration table and uses the branch you're on. No branch names are hardcoded.
Work Item States Follow the Environments
Every environment has a work item state, and the promotion commands keep them in sync. The "Ready for …" states are human sign-offs; the commands never set them.
/deploy-release, /cherry-pick, and /promote check the gate before creating the PR, ask who verifies each product group while the PR is in review, then watch the PR and the environment's CD pipeline — on a green pipeline they set the state and the assignee together. A failed pipeline changes nothing. /track flags anything whose state lags the branch it's on.
Branch Naming
When you run /implement, a branch is automatically created based on the Azure DevOps work item type:
The branch is always created off the current branch — no assumptions are made about which branch you're on.
Note: The Azure DevOps work item type is "Hot Fix" (two words), but the branch prefix and PR label use
hotfix(one word, lowercase).
Code Promotion Flow
Code flows through environments via PRs, never by direct push. Every merge after main deploys:
feature/AB#1234-... ──PR──▸ main ──PR──▸ dev ──PR──▸ test ──PR──▸ staging ──PR──▸ prod
(work branch) (compare, (Dev) (Test) (Staging) (Production)
no deploy)
Development Workflow Guide
This section walks through the complete workflow from picking up a work item to deploying to production.
Step 1: Implement a Work Item
Start by switching to your project's development branch and running /implement:
cd /path/to/your/project
claude
# Start from the compare branch
git checkout main
# Implement the work item
/implement AB#1234
Claude will:
- Set up the ticket's own git worktree next to your project (
../your-project-AB1234). Every later step runs there, so other Claude sessions in the same repo aren't disturbed - Read the work item from Azure DevOps (extracts type, title, acceptance criteria)
- Summarize the work item and ask you to confirm understanding. You can add context or corrections here
- Explore the codebase and present an implementation plan showing which files will be created, modified, or deleted
- Wait for your approval before writing any code
- Create the branch (e.g.,
story/AB#1234-user-can-view-history) and implement using backend and/or frontend agents - Run all quality checks (build, lint, tests, review)
- Generate a UAT checklist and pause for you to manually test
- After you confirm "testing passed", create a PR targeting
main, close the child Task with its hours, and move the work item toCode Review - Remove the worktree once the branch is pushed. It's kept, with a note, if anything is uncommitted or unpushed.
/reworkrecreates it from the branch
Say /implement AB#1234 work in place to skip the worktree and work in the current folder, as before.
Add --opus, --sonnet, or --haiku to run every subagent on one model instead of the default Sonnet split: /implement AB#1234 --opus. /rework and /autopilot take the same flag. The main loop stays on your session model, so switch with /model opus first if you want the whole run on Opus.
The PR merges into main. That merge deploys nothing — main is the compare branch. The work reaches Dev when main is promoted:
/promote main dev
Claude asks who verifies each product group on Dev, waits for Serena to approve the PR and merges it, watches the Dev pipeline, and once it's green moves the work items to Ready for Testing and assigns them to those people. (Dev, Test, and Staging merge themselves this way; a production PR you merge yourself.)
Step 2: Deploy Changes (Quick Commits)
For smaller changes that don't need the full /implement workflow:
/deploy "Fix typo in dashboard header"
This commits, pushes, and triggers the pipeline if you're on an environment branch (dev, test, staging, prod). On main or a feature branch it just pushes — nothing deploys until the work is promoted.
Step 3: Group Work Items into a Release
Once multiple work items are merged to main and smoke-tested on Dev, group them into a release:
/create-release 23
Claude will:
- Ask which work items to include — you can provide:
- Specific IDs:
AB#1234, AB#1235, AB#1236 - A query:
all Ready for Testing user stories - A state filter:
all items tagged sprint-5
- Specific IDs:
- Show you the list and ask for confirmation
- Create a
Release #23iteration in Azure DevOps - Assign all work items to the iteration and tag them with
release-23 - Tell you how to deploy:
/deploy-release 23 test, thenstaging, thenprod
Step 4: Deploy a Release to Test, then Staging
/deploy-release 23 test
Claude will:
- Find all work items tagged
release-23and check each is at the gate for Test (Ready for Testing) - Find their associated commits on
dev - Create a release branch:
release/23-to-test - Cherry-pick all commits for each work item
- Create a PR from
release/23-to-test→test - Link all work items to the PR
- Ask who verifies each product group on Test
- Watch the PR and the Test pipeline; on a green pipeline, move the work items to
Testingand assign them to the approved verifiers
After the PR is reviewed and merged, the Test CD pipeline triggers automatically. QA tests there and sets each passing item to Ready for Staging; then /deploy-release 23 staging repeats the process (gate Ready for Staging, state after merge Staging).
Step 5: Test on Staging
QA and stakeholders test on the Staging environment. If issues are found, fix them with /implement and add the fixes to the release.
Step 6: Deploy a Release to Production
When stakeholders have verified on Staging and set the items to Ready to Deploy:
/deploy-release 23 prod
Same process — gate Ready to Deploy, cherry-picks from staging into a PR targeting prod. You merge this one — Claude auto-merges only into dev, test, and staging. After merge the Production CD pipeline triggers; when it comes back green the items move to Deployed and are assigned to whoever you named for each product group.
Selective Deployment
If staging has 5 user stories but only 3 are Ready to Deploy:
Option A: Create a smaller release
/create-release 24
Include only the 3 ready stories, then /deploy-release 24 prod.
Option B: Cherry-pick specific items
/cherry-pick AB#1234 AB#1235 AB#1236 prod
This cherry-picks just those 3 work items without creating a formal release.
Promoting Without a Release
To promote all code from one branch to the next (no cherry-picking):
/promote main dev
/promote staging prod
This creates a PR from the source branch to the next one in the chain containing everything. main → dev is the everyday first hop; staging → prod is for when everything on staging is approved.
You can also auto-detect the next environment:
/promote
If you're on main, it auto-detects main → dev; on staging, staging → prod.
Hot Fix Workflow
For critical production issues:
- Switch to the production branch:
git checkout prod - Run
/implementwith the Hot Fix work item:/implement AB#9999 - Claude creates a
hotfix/AB#9999-fix-crash-on-submitbranch - Automated checks still run (build, lint, tests, review)
- Manual UAT is skipped — you get an abbreviated confirmation instead
- PR targets
proddirectly with ahotfixlabel — merging it deploys straight to production - Then bring the fix back to
main(a second PR from the same branch, or/cherry-pick AB#9999 main) so the nextmain → devpromotion doesn't overwrite it
Rollback a Deployment
If a deployment causes issues:
Revert specific work items:
/rollback AB#1234 prod
Revert the most recent deployment:
/rollback last staging
Claude will:
- Find the commits to revert
- Create a revert branch (e.g.,
revert/2026-09-06-on-prod) - Run
git reverton each commit - Run pre-flight checks on the reverted code
- Create a PR targeting the environment branch
Merge the PR to deploy the rollback. Work item states are left alone — Claude tells you which items were reverted so you can decide. If the reverted commits are also on main, the next promotion brings them back unless the revert is applied there too.
Code Review
For any open PR:
/review 142
Claude reviews for:
- Clean Architecture boundaries (Domain has no infrastructure dependencies)
- Tenant/organizationId enforcement on all database queries
- Missing unit or integration tests for new code
anytypes in TypeScript (should be properly typed)- Security issues (OWASP Top 10, hardcoded secrets)
- Acceptance criteria coverage from the linked work item
Slash Commands Reference
Hooks
These run automatically — no action needed:
Environment Variables
Each team member sets their own. Never commit these.
Add to ~/.zshrc (Mac) or System Environment Variables (Windows):
# MongoDB (if your project uses MongoDB)
export MONGODB_CONNECTION_STRING="mongodb+srv://user:[email protected]/"
# SQL Server (if your project uses SQL Server or Azure SQL)
export MSSQL_CONNECTION_STRING="Server=localhost;Database=MyDb;User Id=sa;Password=...;"
# PostgreSQL (if your project uses PostgreSQL)
export POSTGRES_CONNECTION_STRING="postgresql://user:password@localhost:5432/mydb"
# Stripe (if your project uses Stripe payments)
export STRIPE_SECRET_KEY="sk_test_..."
# Azure CLI (no env var needed — just log in)
az login
# QA sign-in for /qa and /repro — OPTIONAL, per developer. Leave unset and you sign
# in yourself in the browser window (recommended for your own admin account, since a
# password the agent types lands in the local session transcript). The NAMES come
# from your project's `## Environment URLs` section; _TEST / _DEV / _LOCAL / _STAGING
# versions override the plain one on that environment.
export COMPASS_QA_ADMIN_USERNAME="[email protected]"
export COMPASS_QA_ADMIN_PASSWORD="..."
export COMPASS_QA_ADMIN_PASSWORD_TEST="..." # only if your test password differs
# Microsoft Teams (no env vars needed — uses Microsoft Graph device-code auth;
# first invocation prints a code + URL to sign in)
Setting Up the Teams MCP Server
- Go to Azure Portal > App Registrations
- Click New Registration → name it "Claude Code Teams Bot"
- API Permissions → Add Microsoft Graph:
ChannelMessage.SendChannelMessage.Read.AllChat.ReadWriteTeam.ReadBasic.All
- Certificates & Secrets → Create a new client secret
- Copy the Tenant ID, Client ID, and Client Secret
- Set the three
TEAMS_*environment variables above
Memory System
Claude maintains persistent memory across sessions in ~/.claude/projects/.../memory/. The self-improvement hook prompts Claude to save learnings after each session.
Setting Up Your Memory
After installing, start a Claude Code session and say:
Set up my memory profile. I'm [your name], a [your role] at [your company].
I work on [your projects]. I prefer [your preferences].
Claude creates the initial memory files. Each subsequent session adds to them automatically.
Customization
Adding Custom Agents
Create .claude/agents/your-agent.md in your project:
---
name: your-agent
description: What this agent does
tools:
- Bash
- Read
- Write
---
Instructions for the agent...
Adding Custom Hooks
- Create
.claude/hooks/your-hook.shand make it executable - Add it to
.claude/settings.jsonunder PreToolUse, PostToolUse, or Stop
Adding Custom Slash Commands
Create .claude/commands/your-command.md:
Do something with $ARGUMENTS.
1. Step one
2. Step two
Use it: /your-command some-argument
Removing Components
Delete any agent, hook, command, or MCP server you don't need. Everything works independently.
Updating
When the infrastructure package is updated:
# npx always fetches the latest version
npx @chris1807/claude-kit init /path/to/project
The installer detects existing files and asks whether to overwrite or skip each one.
Publishing Updates (Maintainers Only)
cd claude-kit
# Make your changes, then bump the version
npm version patch # 2.0.0 → 2.0.1
# Push the tag — GitHub Actions publishes to npm automatically
git push && git push --tags
The
.github/workflows/publish.ymlworkflow runs on any pushedv*tag and publishes to npm using theNPM_TOKENrepo secret. You should not runnpm publishfrom a dev machine.
Security
- No secrets in the repo —
.mcp.jsononly contains${ENV_VAR}references - Sensitive data blocker hooks (4 layers) — blocks database queries referencing TIN, SSN, bank account numbers, or other PII fields before execution (Bash + MCP), and scans all output (Bash, MCP, Read, Grep) for PII field names. Even encrypted values are never exposed. CLAUDE.md policy is injected at the top of every project
- Secret blocker hook — automatically blocks writes containing hardcoded credentials
- Protected files hook — prevents edits to production/staging configs
.claude/settings.local.jsonis gitignored — personal permissions stay private- Public npm package — published from a GitHub Actions workflow gated on a tag-version match check
Troubleshooting
npm ERR! 404 Not Found
If npx @chris1807/claude-kit reports 404, your local npm registry may be set to a private feed that overrides the public one. Check with npm config get registry — it should be https://registry.npmjs.org/.
Hooks not running
Verify the settings file is loaded:
claude
/config
Check that .claude/settings.json shows your hooks.
MCP server not connecting
Check that the environment variable is set:
echo $MONGODB_CONNECTION_STRING # Should show your connection string
echo $STRIPE_SECRET_KEY # Should show sk_test_...
az account show # Should show your Azure subscription
Support
- Issues: github.com/Christopher-Waters/claude-kit/issues
- Source: github.com/Christopher-Waters/claude-kit
License
MIT