Skip to content

Built-in skills and plugins

Every project container carries a set of Claude Code resources before you configure anything: skills for code review and site auditing, an output style, a status line, base settings, and a handful of Anthropic plugins installed by default. This page lists what is actually in the box.

At container start, entrypoint.sh symlinks each file and skill from the resource bundle into ~/.claude/, entry by entry. Files your team commits to its own .claude/ directory take precedence over the bundled ones. Speedwave tracks which links it owns in ~/.claude/.speedwave-managed-links, so a restart cleans up links for anything you have since disabled.

Fourteen skills each check one thing: basic, change-impact, comment-analyzer, documentation-checker, duplication-detector, kiss-detector, performance-concurrency, security-checker, silent-failure-hunter, solid-detector, ssot-detector, test-analyzer, type-design-analyzer, and yagni-detector.

speedwave-code-review is the aggregate skill and the deepest review Speedwave ships: invoked as /speedwave-code-review, it runs all fourteen in parallel over a changeset and merges only the verified findings. It is a different tool than Claude Code’s built-in /code-review command, which covers everyday diff reviews and drives pull and merge request review through the integrations, the flow described in AI-assisted code review.

Two more skills work on websites rather than code:

SkillWhat it does
speedwave-site-auditAudits a URL against a bundled checklist covering foundations, SEO, accessibility, security, performance, privacy, resilience, i18n, well-known URIs, and agent readiness. Takes a URL, an optional --scope page|sitemap, a category filter, and an output path.
speedwave-sitemapBuilds a JSON inventory of a site’s URLs. It reads robots.txt and sitemap.xml first, falling back to a bounded, same-domain Playwright crawl.

There is also speedwave-product-showcase, which builds a self-contained, animated product demo for a landing page.

skills/integrations/<service>/ holds a skill for each of 13 services: Atlassian, Calendar, Context7, GitHub, GitLab, Mail, Notes, Office, Playwright, Redmine, Reminders, SharePoint, and Slack. Only the skills for services you enable get linked in, the same gating that controls which integrations are available.

Every container also ships an output style (Speedwave.md), a status line script (statusline.sh), and a base settings.json with effortLevel: high, outputStyle: Speedwave, and showThinkingSummaries: true. These give Claude a consistent voice and a visible thinking summary without any setup on your part.

At container start, Speedwave also installs five official Anthropic plugins from the claude-plugins-official marketplace: frontend-design, feature-dev, claude-md-management, superpowers, and typescript-lsp. The install is idempotent and non-fatal. Disable one with /plugin. That choice persists across restarts; it is not reinstalled for you.

These are separate from Speednet’s signed plugins, which are written, developed, and shipped exclusively by Speednet: the bundled Anthropic plugins are unsigned, come from Anthropic’s own marketplace, and cannot be turned off per project the way a signed plugin can. For how the runtime assembles all of this at startup, see how Speedwave works.

A session-start hook injects using-superpowers on every session, including after /clear or compaction, so Speedwave brainstorms a design, writes a plan, executes it with reviewed subagents, and verifies before calling anything done. See your first session for what that looks like in your first chat, and the default development workflow for the full stage-by-stage reference. Disable it with /plugin disable superpowers; the choice survives a restart.