Install the engineering plugin and get ten skills mapped to the working engineer's day

A plugin in the Claude Code / Cowork sense is a bundle of skills, agents, slash commands, and MCP server entries you install as one unit. The engineering plugin, published in the Anthropic marketplace, packages ten skills that cover the workdays of a software engineer or SRE: architecture for ADRs, system-design for new services, code-review for diff review, debug for structured triage, deploy-checklist for pre-flight, incident-response for sev calls and postmortems, standup for daily updates, tech-debt for refactor backlogs, testing-strategy for test plans, and documentation for runbooks and READMEs. Installing the plugin lands all ten in ~/.claude/plugins/; only the descriptions stay in context, and a skill's body loads only when its description matches what you've asked for.

Why pick this as the first plugin to feature. Plugins are how the Claude stack scales past one-off skills — instead of installing ten skills by hand and keeping them in sync with the upstream, you install one plugin and updates ride along. The engineering plugin is the canonical reference: a coherent set of skills authored against one voice, with overlapping vocabularies that mean a single prompt ("this PR is making me nervous") can route to code-review or debug depending on whether you've pasted a diff or a stack trace. It is also the lowest-risk install for an engineering-leaning user, since every skill in the bundle has a clear failure mode that does no harm if it triggers wrongly — a code review on a non-PR is just a list of observations.

Pre-flight work. Before code ships, the bundle gives you three skills that map to the three planning conversations you already have. architecture writes or evaluates an ADR — useful when you are choosing between technologies (Kafka vs SQS, Postgres vs DynamoDB) and want the trade-off table written down somewhere a future engineer will actually find it. system-design takes requirements and constraints and proposes a service topology, an API shape, and a data model. testing-strategy turns "how do we test this?" into a concrete plan covering unit, integration, and contract tests, with a stated coverage target. Each of these is the kind of prompt you would otherwise type from scratch every quarter.

In-flight work. Once code is moving, three more skills cover the moments that actually decide whether a change ships well. code-review looks at a diff or PR URL for security, performance, and correctness issues — N+1 queries, missing error paths, injection risks, edge cases the author missed. debug runs a structured reproduce-isolate-diagnose-fix session against a stack trace or a "works in staging but not prod" report. deploy-checklist is the one to wire into a pre-deploy hook: it verifies migrations, feature flags, CI status, and asks for rollback triggers before you cut the release. Together these three are most of what a senior engineer spends a Tuesday on.

Post-flight work. The last four skills cover the work around the work. incident-response triages an alert, drafts the customer-facing status update, and writes the blameless postmortem when it's over. standup turns yesterday's commits, PRs, and ticket moves into a yesterday/today/blockers update you can paste into Slack. tech-debt audits a codebase or directory for refactor candidates and helps you prioritize them against feature work. documentation writes the README, the runbook, or the API doc you keep meaning to write and never do.

The trap to avoid is treating an installed plugin as ten always-on assistants. None of these skills are loaded by default — they trigger on description match, which means your prompt needs to use language the description recognises. If code-review isn't firing, paste a diff or say "review this before I merge." If incident-response isn't firing, say "we have an incident" or paste the alert. Skills that never trigger are dead weight; the fix is almost always rephrasing the prompt, not editing the skill.

See the try-it block for the install and a one-line probe that confirms a skill triggers.