First steps

With the plugins installed, there are three things to do: set up the repository, learn how components are invoked, and run a first task.

Set up the repository with /tsh-core:init

Run it once per repository, inside that repository:

/tsh-core:init

init takes a repository from any starting state, whether nothing, a partial setup or a drifted one, to a complete set of Claude Code context. It audits what already loads, then creates or repairs:

• the root CLAUDE.md, and nested ones where ownership demands,
• path-scoped rules in .claude/rules/,
• the decision-record archive and its index.

It does this by running the skills that own those artifacts, in order, and it adds a maintenance section to CLAUDE.md so future sessions know which skill keeps each artifact true.

• It is safe to re-run: the second pass is a repair, not a rewrite.
• It never overwrites an existing CLAUDE.md wholesale; it revises in place and reports what changed.
• It does not invent decision records; it lists candidates for a human to confirm.
• It ends with a report of what was created, repaired and left alone, and flags the result as a proposal for your team to review.

init is a command you type. It is deliberately not model-invocable, so a setup request phrased in prose routes to managing-claude-context instead.

Invoke skills and agents

Both are namespaced by plugin, and they are invoked differently:

/tsh-product-testing:<skill-name>     # skills: type / and the plugin name
@tsh-product-testing:<agent-name>     # agents: @-mention typeahead

• Most skills are model-invocable: Claude loads one when your request matches its description, so you can describe the work ("implement this ticket", "is this ready to merge") instead of remembering a command.
• Some skills are commands only. /tsh-core:init is one; in tsh-product-management the entry points analyze-materials and explore-materials are too.
• Some skills are internal contracts that agents preload and stay out of the / menu, for example capturing-ui-evidence.

To see exactly what a plugin gives you, run /plugin, open the Installed tab and press Enter on it, or run claude plugin details tsh-product-testing@tsh-agentic-collections from the shell.

Run a first task

Pick the line that matches your work. Each links to the full page.

You want to	Try	Needs
Implement a feature, ticket or plan end to end	/tsh-product-engineering:orchestrating-feature-implementation	tsh-product-engineering
Break a feature into a committed plan file	/tsh-product-engineering:creating-implementation-plans	tsh-product-engineering
Fix a bug at its root cause	/tsh-product-engineering:debugging-code	tsh-product-engineering
Review a change before merge	/tsh-product-engineering:reviewing-code	tsh-product-engineering
Check a UI against its Figma design	/tsh-product-engineering:reviewing-ui	Figma sign-in, playwright-cli, a running app and its exact dev server URL
Write or de-flake Playwright E2E tests	/tsh-product-testing:writing-playwright-e2e-tests or @tsh-product-testing:e2e-engineer	tsh-product-testing
Turn workshop material into a Jira-ready backlog	/tsh-product-management:analyze-materials with a transcript path, a Figma or FigJam link, a PDF or a folder	Atlassian sign-in for anything Jira

For the end-to-end flows, see the workflows.

If something does not show up

See Troubleshooting.