v1.2.2

v1.2.2 — Fix three skill descriptions

A small fix that completes v1.2.1. No change to what the commands do.

What was wrong

After v1.2.1, flow-test still showed a second entry in the slash menu, described as "Flow Test (web, in the browser)".

The description line of three skills contained a colon followed by a space:

• flow-test: "…without a spec: for a bug fix…"
• feature-slice: "Two modes: scaffold…"
• zustand-slice: "Two modes: scaffold…"

That is not valid in a skill's header, so Claude Code could not read the header of those three files at all. As a result, the setting added in v1.2.1 to hide them from the slash menu was ignored, and their descriptions were not being read either.

What changed

The three descriptions now use a dash instead of the colon. Nothing else in the skills was touched.

	Before	After
/flow-test in the menu	An extra "Flow Test (web, in the browser)" entry	One entry
Running the commands	Works	Works the same
Claude reading these three skills' descriptions	Not read	Read

Because the descriptions are readable again, Claude can now pick up flow-test, feature-slice and zustand-slice on its own from what you ask for, without you typing the command.

All skill, command and agent files in the kit were checked for the same problem. No others have it.

Upgrading

1. In Claude Code: /plugin marketplace update agentic-dev-kit
2. Then /reload-plugins, or restart Claude Code.

Nothing needs to change in your projects.

Full Changelog: v1.2.1...v1.2.2

v1.2.1 - One slash-menu entry per command

v1.2.1 — One slash-menu entry per command

A small fix for the slash menu. No behaviour changes.

What was wrong

Six features showed up twice when you typed / in Claude Code: flow-test, docs-sync, ship, spec-issues, feature-slice and zustand-slice. Each one has a command and a skill with the same name, and the menu listed both. The two entries did the same thing, so it wasn't clear which to pick.

What changed

Those six skills are now hidden from the slash menu. The command is the single entry you see and type.

	Before	After
/flow- in the menu	flow-test listed twice	flow-test listed once
Running /agentic-dev-kit:flow-test	Works	Works the same
Claude loading the skill on its own	Works	Works the same

The same applies to the other five.

Upgrading

1. In Claude Code: /plugin marketplace update agentic-dev-kit
2. Restart Claude Code.

Nothing needs to change in your projects, and there is no need to rerun /agentic-dev-kit:init-dev-kit for this release.

Full Changelog: v1.2.0...v1.2.1

v1.2.0

v1.2.0 — From spec to pull request

This release adds a complete workflow around the existing build skills. Until now the kit helped developers build (tokens, feature slices, components, design QA, native port). It now also covers what comes before and after: writing down how a flow should behave, turning that into GitHub issues, and checking the work before it is merged.

The workflow

/spec-builder → /spec-challenger → PM approves → /spec-issues → build → /ship

/ship runs lint and type-check, /design-qa, /pr-check, /security-check, /flow-test and /docs-sync in order, stops on a blocker, and opens the pull request. Merging stays a human decision.

New commands

Command	Who runs it	What it does
/agentic-dev-kit:spec-builder	PM	Interviews the PM and writes a flow's functional spec to docs/specs/<flow-name>/spec.md.
/agentic-dev-kit:spec-challenger	Developer	Challenges the spec before approval and raises the doubts a developer would hit mid-build.
/agentic-dev-kit:spec-issues	PM or lead	Turns an approved spec's work breakdown into a GitHub epic and child issues.
/agentic-dev-kit:pr-check	Developer	Reviews a branch against the spec and the rules in CLAUDE.md.
/agentic-dev-kit:security-check	Developer	Reviews the frontend changes for security problems.
/agentic-dev-kit:flow-test	Developer	Checks acceptance criteria on the running web app, in the browser.
/agentic-dev-kit:docs-sync	Developer	Brings the spec, feature doc, changelog and README in line with what was built.
/agentic-dev-kit:ship	Developer	Runs the checks in order and opens the PR with one combined report.

What each one does

Functional specs

• /spec-builder asks questions in rounds (why, scope, the flow screen by screen, branches, APIs, offline, errors, sensitive data) and never invents an answer. Anything unanswered becomes an open question with an owner.
• The spec has 16 sections. Screens, steps, rules, APIs, acceptance criteria, open questions and work items each get a permanent ID (S-, F-, R-, API-, AC-, Q-, W-), so later work can point at them.
• When a Figma link is given, the screen list is read from the Figma file.
• A spec stays Draft while a blocking question is open. Only the PM's confirmation makes it Approved.
• /spec-challenger is a read-only agent that checks the spec against itself, the Figma file and the codebase. Doubts the developer agrees with are added to the spec's open questions, tagged [challenge].

Issues

• /spec-issues works on Approved specs only. It creates one epic and one child issue per work item, in dependency order, and writes the issue numbers back into the spec.
• Issues follow the project's feature-request issue template in .github/ISSUE_TEMPLATE/. Titles are feat(<flow-name>:W-01): <piece of work> and feat(<flow-name>:epic): <flow name>.
• It shows the full plan and asks once before creating anything, and it does not create duplicates when run again.

Checks

• /pr-check and /security-check use read-only agents. They report first, then fix only what the developer agrees to. Reports are posted to the PR only when asked.
• /flow-test drives the developer's own Chrome through the Claude in Chrome extension. It reports pass, fail or could-not-test per acceptance criterion, with console and network evidence, and saves rerunnable test files in the project's existing end-to-end runner.
• /docs-sync shows every planned document edit and writes only what the developer agrees to. Where the code differs from the spec, it raises a [drift] question for the PM instead of rewriting the spec.

Works without a spec too

For a bug fix or small change raised straight as an issue, /pr-check, /security-check, /docs-sync and /ship run on the branch's changes and skip only their spec-based checks. /flow-test #<issue> tests against the issue's stated expected behaviour. Only /spec-issues needs a spec.

Fixes

• Skills now use the @app/* and @8848digital/* workspace package scopes consistently.
• feature-slice imports chassis symbols from @8848digital/catalyst.

Upgrading

1. Update the plugin in Claude Code: /plugin marketplace update agentic-dev-kit, then restart.
2. In each project, run /agentic-dev-kit:init-dev-kit and choose merge, to bring the new guardrails into the project's CLAUDE.md.

What you need

• GitHub CLI (gh), signed in, for /spec-issues, /ship and posting reports to a PR.
• Figma MCP connected, for reading screens in /spec-builder and /spec-challenger.
• Claude in Chrome extension, for /flow-test.

Known limitations

• /flow-test covers the web app only. Native testing stays manual.
• The checks do not replace human review. Sync and offline behaviour, SQL correctness and behaviour on real devices still need a person.

Full Changelog: v1.1.0...v1.2.0

v1.1.0

Second release of agentic-dev-kit. Adds Zustand global-state tooling and a bare React Native CLI skill, bringing the kit to 10 skills.

What's new since v1.0.0

New skills

• zustand-slice — Zustand global-state slice patterns (store anatomy, selectors, persist/storage)
• react-native-expert — bare React Native CLI expertise adapted to the kit's MO (navigation, list optimization, platform handling, storage hooks, project structure)

New command

• zustand-slice

Full contents (10 skills)

web-component, rn-component, feature-slice, design-system-setup, web-to-native, nextjs-nerd, react-renderer, typescript-teacher, zustand-slice, react-native-expert — plus the design-qa agent and the init-dev-kit / design-qa / feature-slice / zustand-slice commands.

Getting started

After enabling the plugin, run:

/agentic-dev-kit:init-dev-kit

The skills are coupled by design — adopt them together, not cherry-picked.

Full changelog: v1.0.0...v1.1.0

v1.0.0

First release of agentic-dev-kit — an opinionated Claude Code plugin for agentic frontend development on a Turborepo + pnpm / Next.js web / bare React Native CLI / Figma MCP / TanStack Query / Frappe stack.

What's included

8 skills

• web-component — web UI components
• rn-component — React Native components
• feature-slice — feature-local slices
• design-system-setup — design-system tokens
• web-to-native — web → native migration
• nextjs-nerd — Next.js App Router / server components & actions
• react-renderer — React 19 patterns & performance
• typescript-teacher — TypeScript domain typing

1 agent

• design-qa — read-only design QA

3 commands

• init-dev-kit — installs the CLAUDE.md guardrails into your project
• design-qa
• feature-slice

Getting started

After enabling the plugin, run:

/agentic-dev-kit:init-dev-kit

The skills are coupled by design — adopt them together, not cherry-picked.

Note: a further set of general-purpose skills (testing, security, language pros) is planned for a later release.