crew

Community best practices for working with Claude Code - context management, planning, TDD, hooks, skills, subagents, slash commands and MCP - with what crew already does, where crew departs on purpose, and what is genuinely unsettled. Use when asked how to use Claude Code well, whether a practice is worth adopting, why crew is built the way it is, or when auditing a repo's CLAUDE.md, hooks or skill layout against community consensus.

The change-request template, the ten questions a change board requires, and how each field maps onto ServiceDesk Plus, Jira and a local file. Use when filing, checking or closing a change request, when a production promotion asks for an approved change, or when the user says change request, change management, CAB, change window, rollback plan or post-change validation.

Configure the official AWS and Azure MCP servers safely for a repo. Use when the user says set up AWS MCP, add the Azure MCP server, connect to my cloud account, look up AWS docs or pricing, or asks how to give the crew read access to cloud resources.

Manage context exhaustion - warn near the limit, write handoff notes, and resume work after a clear or compact. Use when the user says set up handoff, context is filling up, wrap up the session, write a handoff note, resume where we left off, or asks how to clear the session without losing state.

Find the cause of a defect before proposing a fix - read errors, reproduce, check recent changes, gather evidence from this repo's code map and verification map, then form one hypothesis and test it minimally. Use when encountering any bug, test failure, flaky test, performance problem or unexpected behaviour, before writing a fix. Also use when a previous fix did not work, when three fixes have failed, or when time pressure makes guessing tempting.

Author architecture, process, sequence, ER and data-flow diagrams as Mermaid, render them to PNG or SVG, and produce Visio files when Visio is installed. Use when the user says draw a diagram, make an architecture diagram, show the data flow, diagram this process, export to PNG, or asks for a Visio version.

Keep CHANGELOG.md, README.md, SECURITY.md, TODO.md, ADRs, generated docs, and the API and feature reference current as work lands. Use when the user says update the changelog, keep the docs in sync, update the README, maintain TODO, document the API, list the endpoints or features, or asks which docs a change should touch.

Build and query a code graph of this repository with graphify, keeping it on the filesystem as graph.json and fresh on commit. Use when the user asks to build or refresh the code graph, asks what calls what or what connects to what, or asks how a subsystem is wired. The graph lives on the filesystem only; there is no vault or notes-app export and no config for one.

House style for documents the crew hands to a human - palette, heading hierarchy, capitalization, and whether the artifact should be PDF, DOCX, HTML or plain markdown. Use when exporting an architecture write-up, runbook, report or handoff, when asked what a delivered document should look like, or when choosing between PDF and DOCX.

Configure and run linters and formatters for PowerShell, PHP, Python, Terraform, and JavaScript, and wire them into the verification gate. Use when the user says set up linting, add a linter, format this code, run PSScriptAnalyzer, ruff, phpstan, or tflint, or asks why style keeps drifting.

Conventions for durable project memory across repos, including Obsidian vault integration. Use when the user says set up memory, wire up Obsidian, connect my vault, or when writing or reading code maps, decision notes, or cross-repo contracts.

Set up outbound notifications to Microsoft Teams or Telegram, and explain the two-way MCP options. Use when the user says set up notifications, notify me, send to Teams, set up a Telegram bot, post updates to a channel, wire up a webhook, or asks how to get alerted when a phase finishes or a gate fails.

Manage the crew itself - report crew status, decide which roles the crew should have, onboard a repo or a role, offboard a role, and keep session context and the code map from going stale. Use when the user asks about crew status, who is on the crew, whether the crew is the right size, onboarding or offboarding a role or repo, or says the map is out of date. Not for general "how do I" questions.

Set up, verify, and invoke external model providers - Codex or GitHub Copilot for QA review, Gemini for design second opinions. Use when the user says set up codex, set up copilot, set up gemini, add a reviewer, add a design partner, configure providers, wire up the API key, or asks about free tiers, rate limits, model families, which model to use for a role, or why a provider call is failing.

Write, index, and maintain operational runbooks - deploys, rollbacks, alert response, recurring jobs, disaster recovery. Use when the user says write a runbook, document the deploy process, what do I do when X alerts, how do we roll this back, or asks to capture an operational procedure for later.

Decide when to add or remove agent roles and when to parallelize. Use when the user says grow the team, expand the crew, add more agents, run agents in parallel, speed this up, increase throughput, or asks whether the setup is worth its cost or the right size.

Bootstrap and configure a repository for the crew workflow, as guided phases. Use whenever the user says any of - set up crew, set up the crew, set up the team, set up the virtual team, phase setup, phased setup, run the setup, initialize crew, init crew, configure crew, onboard this repo, get this repo ready, add crew to this project, start the crew here - or otherwise asks to install, configure, or resume crew setup in a repository. Also use when they ask where setup got to, what phase they are on, or what is left to configure.

Set up and run terraform-docs and tflint for a Terraform module, including the header comment block, footer.md, and README injection. Use when the user says set up terraform-docs, generate module docs, add tflint, lint terraform, or asks why the Terraform README keeps changing or reverting.

Build the verification map, wire test credentials safely, and set up Playwright. Use when the user asks which tests should run after a change, says set up the test map, add browser tests, add visual testing, validate UI changes, or when tests need credentials from AWS Secrets Manager, Vault, Key Vault, or a .env file.

Discover and install agent skills from the open skills ecosystem via the skills CLI. Use only when the user explicitly asks to find, search for, browse, or install a skill or plugin - not for general "how do I" questions, which should be answered directly.