Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Fix stale shipped kit docs and counts, and bump kit versions
Co-authored-by: Atlas <AtlasScope@users.noreply.github.com>
  • Loading branch information
cursoragent and AtlasScope committed Nov 19, 2026
commit 217dc2fa24f7e631a9e2a01634a2412b1e5c7079
6 changes: 3 additions & 3 deletions .claude-plugin/marketplace.json
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@
{
"name": "design-kit",
"description": "A design library for AI coding agents: a /design router that picks the right skill or recipe, plus vetted skills for typography, color, motion, UX writing, usability, landing copy, native iOS and more.",
"version": "1.4.0",
"version": "1.4.1",
"source": "./kits/design-kit/plugins/design-kit",
"category": "design",
"homepage": "https://github.com/AtlasScope/agent-kit/tree/main/kits/design-kit",
Expand Down Expand Up @@ -44,7 +44,7 @@
{
"name": "engineering-kit",
"description": "Engineering skills for AI coding agents, vetted and pinned: bug diagnosis, test-first work, architecture scans, plan interviews, retros, primary-source research, setup wizards and read-only PostgreSQL investigations. Every command runs only when you call it, unless you choose at install to let /engineer and /database start on their own.",
"version": "1.6.0",
"version": "1.6.1",
"source": "./kits/engineering-kit/plugins/engineering-kit",
"category": "engineering",
"homepage": "https://github.com/AtlasScope/agent-kit/tree/main/kits/engineering-kit",
Expand All @@ -59,7 +59,7 @@
{
"name": "experience-architecture",
"description": "Find out why a web app's navigation, pages, links or Back button feel wrong, decide whether its structure should change, and prove it in a real browser. Diagnose, Rethink and Verify over a zero-dependency toolkit for route maps, browser journeys, prototypes and URL migrations, plus Walkthrough (preview), which learns the product by using it, says what it should become and can carry the improvement through. Strongest on React and Next.js.",
"version": "0.4.0",
"version": "0.4.1",
"source": "./kits/experience-architecture/plugins/experience-architecture",
"category": "engineering",
"homepage": "https://github.com/AtlasScope/agent-kit/tree/main/kits/experience-architecture",
Expand Down
2 changes: 1 addition & 1 deletion kits/design-kit/claude-ai/SKILL.md
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@ description: A whole design library in one skill. Use for any interface design o

# Design Kit

This one skill holds a full design library: about thirty specialist skills, twenty-two recipes that chain them, and the rules for choosing between them. The user does not want to pick a skill. Pick for them, say which in one line, and do the work.
This one skill holds a full design library: 36 specialist skills, 23 recipes that chain them, and the rules for choosing between them. The user does not want to pick a skill. Pick for them, say which in one line, and do the work.

## How to choose

Expand Down
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"name": "design-kit",
"description": "A design library for AI coding agents: a /design router that picks the right skill or recipe, plus vetted skills for typography, color, motion, UX writing, usability, landing copy, native iOS and more.",
"version": "1.4.0",
"version": "1.4.1",
"author": {
"name": "AtlasScope"
},
Expand Down
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"name": "design-kit",
"description": "A design library for AI coding agents: a /design router that picks the right skill or recipe, plus vetted skills for typography, color, motion, UX writing, usability, landing copy, native iOS and more.",
"version": "1.4.0",
"version": "1.4.1",
"author": {
"name": "AtlasScope"
},
Expand Down
10 changes: 5 additions & 5 deletions kits/design-kit/plugins/design-kit/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@

A design library for AI coding agents. Type `/design` and what you want; the router picks the one right skill or recipe from the library, says why in a line, and starts it. Recipes run one step at a time and resume with `/design next`.

It bundles 38 skills: the router, `/design-setup` (which explains and saves the two choices Claude Code asks about at install), Emil Kowalski's motion and polish skills, Hallmark, wondelai's UX design set (typography, Refactoring UI, heuristics, microinteractions, iOS HIG and more), two color skills, UX writing, landing-page copy and conversion, SwiftUI Pro, Stop Slop and fix-writing. Six of them (`apple-design`, `break-ui`, `color-expert`, `fix-writing`, `mobile-native`, `stop-slop`) are listed to the agent in every session, so it can reach for them unprompted. The rest stay out of that list to keep sessions light: they run when you type their name or when `/design` picks them. Credits, versions and licenses are in [THIRD_PARTY.md](https://github.com/AtlasScope/agent-kit/blob/main/kits/design-kit/THIRD_PARTY.md).
It bundles 38 skills: the router, `/design-setup` (which explains and saves the three choices Claude Code asks about at install), Emil Kowalski's motion and polish skills, Hallmark, wondelai's UX design set (typography, Refactoring UI, heuristics, microinteractions, iOS HIG, and more), two color skills, UX writing, landing-page copy and conversion, SwiftUI Pro, Stop Slop, and fix-writing. Six of them (`apple-design`, `break-ui`, `color-expert`, `fix-writing`, `mobile-native`, `stop-slop`) are listed to the agent in every session, so it can reach for them unprompted. The rest stay out of that list to keep sessions light: they run when you type their name or when `/design` picks them. Credits, versions, and licenses are in [THIRD_PARTY.md](https://github.com/AtlasScope/agent-kit/blob/main/kits/design-kit/THIRD_PARTY.md).

The router also routes to skills that ship as their own plugins. Install the ones you want alongside this kit:

Expand Down Expand Up @@ -52,13 +52,13 @@ Cloud sessions only get what the repo declares. Add this to the repo's `.claude/

**claude.ai chats and Projects**

Build one zip that holds the whole library as a single skill, then upload it once per account at Settings > Capabilities > Skills (code execution must be on):
Build one zip that holds the whole library as a single skill, then upload it once per account at Settings > Capabilities > Skills (code execution must be on). Run from `kits/design-kit`:

```sh
scripts/build-claude-ai-zips.sh
```

The zip is `dist/claude-ai/design-kit.zip`. Its own instructions (`claude-ai/SKILL.md`) pick the right specialist and read it from inside the bundle. To upload single skills instead, name them: `scripts/build-claude-ai-zips.sh oklch-skill ux-writing`.
The zip is written to `kits/design-kit/dist/claude-ai/design-kit.zip`. Its own instructions (`claude-ai/SKILL.md`) pick the right specialist and read it from inside the bundle. To upload single skills instead, name them: `scripts/build-claude-ai-zips.sh oklch-skill ux-writing`.

Skills uploaded there run in claude.ai's sandbox: they advise and write, but cannot reach your repos.

Expand All @@ -80,6 +80,6 @@ The router's references (`catalog.md`, `recipes.md`, `situations.md`) are genera
node kits/design-kit/scripts/build_references.mjs
```

`node library/check.mjs` fails when they are out of date. The Impeccable helper agents are not library entries; their rows live in `scripts/impeccable-helpers.json`.
`node library/check.mjs` fails when they are out of date. The Impeccable helper agents are not library entries; their rows live in `kits/design-kit/scripts/impeccable-helpers.json`.

Third-party skills change only by re-reading the new upstream version and committing it with its new commit in THIRD_PARTY.md.
Borrowed skills change only by re-reading the new upstream version and committing it with its new commit in `THIRD_PARTY.md`.
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"name": "engineering-kit",
"description": "Engineering skills for AI coding agents, vetted and pinned: bug diagnosis, test-first work, architecture scans, plan interviews, retros, primary-source research, setup wizards and read-only PostgreSQL investigations. Every command runs only when you call it, unless you choose at install to let /engineer and /database start on their own.",
"version": "1.6.0",
"version": "1.6.1",
"author": {
"name": "AtlasScope"
},
Expand Down
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"name": "engineering-kit",
"description": "Engineering skills for AI coding agents, vetted and pinned: bug diagnosis, test-first work, architecture scans, plan interviews, retros, primary-source research, setup wizards and read-only PostgreSQL investigations. Every command runs only when you call it, unless you choose at install to let /engineer and /database start on their own.",
"version": "1.6.0",
"version": "1.6.1",
"author": {
"name": "AtlasScope"
},
Expand Down
6 changes: 3 additions & 3 deletions kits/engineering-kit/plugins/engineering-kit/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ It is the engineering sibling of [design-kit](https://github.com/AtlasScope/agen

| Command | What it does |
|---|---|
| `/engineer` | Describe what is going on and get one starting move: an investigation prompt to paste, the skill to run, or a thinking prompt. It suggests and stops; it runs nothing until you say go. Catalog from the Agent Engineering Field Guide. |
| `/engineer` | Describe what is going on and get one starting move: an investigation prompt to paste, the skill to run, or a thinking prompt. It suggests and stops; it runs nothing until you say go. Catalog from `library/library.json` (originally the Agent Engineering Field Guide). |
| `/database` | Describe a PostgreSQL problem in plain words (wrong numbers, heavy writes, a crowded schema, cost, a risky migration) and get one database investigation, run read-only when the session can reach the code and database. Works on any Postgres host. Prompts come from the prompt library. |
| `/diagnosing-bugs` | Hard bugs and performance regressions: build a loop that goes red, minimise, rank hypotheses, instrument, fix with a regression test, clean up. |
| `/tdd` | Test-first work in vertical slices, with the seams agreed before any test is written. |
Expand All @@ -20,7 +20,7 @@ It is the engineering sibling of [design-kit](https://github.com/AtlasScope/agen
| `/wizard` | Writes a bash script that walks you through steps only a human can do (dashboards, secrets, cutovers). |
| `/to-questionnaire` | Turns a decision you cannot answer alone into a questionnaire for the person who can. |
| `/wait-what` | Asks the agent to re-pitch its last message in plain language. |
| `/engineering-setup` | Claude Code only. Explains in plain words whether `/engineer` and `/database` should start on their own, then saves your answer to your settings after showing the change. Takes effect next session. |
| `/engineering-setup` | Claude Code only. Explains in plain words the two install choices (whether `/engineer` and `/database` start on their own, and whether to keep a local log of picks), then saves your answers to your settings after showing the change. Takes effect next session. |

## Pairs well with

Expand Down Expand Up @@ -78,7 +78,7 @@ The router's references (`start-here.md` and `investigations.md`) come from `lib

## Rebuilding the /database reference

The `/database` prompts come from `library/library.json`. After editing them there, run `node kits/engineering-kit/scripts/build_database_references.mjs` from the repo root.
The `/database` prompts come from `library/library.json`. After editing them there, run `node kits/engineering-kit/scripts/build_database_references.mjs` from the repo root. `node library/check.mjs` fails while it is out of date.

## Updating a vendored skill

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -9,12 +9,12 @@ metadata:

# Engineer

The user types `/engineer` and what is going on, in their own words, or just describes it if they chose automatic start. They have 50 investigations, 10 thinking prompts and two dozen skills, and do not want to remember which fits. They get one starting move and one line on why. This router suggests; it does not run a workflow.
The user types `/engineer` and what is going on, in their own words, or just describes it if they chose automatic start. They have 56 investigations, 12 thinking prompts and 30 skills, and do not want to remember which fits. They get one starting move and one line on why. This router suggests; it does not run a workflow.

The library is in two references, generated from the guide:

- `references/start-here.md`: symptoms mapped to investigations, every skill with the investigations it covers, the thinking prompts in full, and how to wrap an investigation prompt. Read it first, every time.
- `references/investigations.md`: all 50 investigations with the prompt, what it finds, the evidence to ask for, when to skip it, and the usual agent blind spot. Read only the entries you are weighing.
- `references/investigations.md`: all 56 investigations with the prompt, what it finds, the evidence to ask for, when to skip it, and the usual agent blind spot. Read only the entries you are weighing.

## Settings

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,7 @@ metadata:

# Engineering setup

The user typed `/engineering-setup`. Explain the choice so someone who has never seen the kit can answer, ask, then save the answer with one yes. Claude Code also asks this question when the kit is turned on; this is the friendlier way through the same choice, and it can be run again any time.
The user typed `/engineering-setup`. Explain both choices so someone who has never seen the kit can answer, ask, then save the answers with one yes. Claude Code also asks these questions when the kit is turned on; this is the friendlier way through the same choices, and it can be run again any time.

## Where this works

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
"$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json",
"name": "experience-architecture",
"displayName": "Experience Architecture",
"version": "0.4.0",
"version": "0.4.1",
"description": "Find out why a web app's navigation, pages, links or Back button feel wrong, decide whether its structure should change, and prove it in a real browser. Diagnose, Rethink and Verify over a zero-dependency toolkit for route maps, browser journeys, prototypes and URL migrations, plus Walkthrough (preview), which learns the product by using it, says what it should become and can carry the improvement through. Strongest on React and Next.js.",
"author": {
"name": "AtlasScope"
Expand Down
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"name": "experience-architecture",
"description": "Find out why a web app's navigation, pages, links or Back button feel wrong, decide whether its structure should change, and prove it in a real browser. Diagnose, Rethink and Verify over a zero-dependency toolkit for route maps, browser journeys, prototypes and URL migrations, plus Walkthrough (preview), which learns the product by using it, says what it should become and can carry the improvement through. Strongest on React and Next.js.",
"version": "0.4.0",
"version": "0.4.1",
"author": {
"name": "AtlasScope"
},
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,7 @@ node agent-kit/kits/experience-architecture/plugins/experience-architecture/scri

It installs `/experience-architecture-diagnose`, `-rethink`, `-verify` and `-walkthrough`, and keeps the plugin's files in `~/.experience-architecture/plugin`. Rerun it after a `git pull` to update; add `--uninstall` to remove only what it installed. `--codex` writes the same folders to `~/.codex/skills` if you prefer that to the Codex plugin, and `--dir <path>` targets any other tool that reads skill folders.

To try it in Claude Code for one session without installing: `claude --plugin-dir <clone>/plugins/experience-architecture`.
To try it in Claude Code for one session without installing: `claude --plugin-dir <clone>/kits/experience-architecture/plugins/experience-architecture`.

## What it does

Expand Down
Loading