# assistive-ux

Adopt this skill: https://proskills.md/skills/other/block-berd-agents-skills-assistive-ux

Page: https://proskills.md/skills/other/block-berd-agents-skills-assistive-ux
This document is the public skill bundle for that page. It contains SKILL.md and every helper file the catalog carries. No login is required.

## File: .agents/skills/assistive-ux/SKILL.md
size: 2822
upstream: https://github.com/block/berd/blob/main/.agents/skills/assistive-ux/SKILL.md

````text
---
name: assistive-ux
description: Use when adding, reviewing, designing, testing, or managing Assistive UX moments in Berd, including discover, suggest, and autoApply guidance, adaptive settings, behavior signals, retirement rules, or changes under src/shared/assistive-ux.
---

# Assistive UX

Use this skill for Berd product intelligence moments that teach,
suggest, or adapt settings based on user behavior.

Before changing code, read `docs/assistive-ux.md`. Treat that doc as the source
of product intent and architecture.

## Core Rule

Keep assistive state separate from user settings:

```text
User settings answer: what should the product do?
Assistive UX answers: what has the app shown, observed, accepted, retired, or applied?
```

Do not store guidance lifecycle state inside a feature setting. Do not use
assistive state as the source of truth for product behavior.

## Types

- `discover`: teaches that a control, setting, or capability exists. It should
  not change settings.
- `suggest`: notices a repeated behavior and asks before changing a setting.
- `autoApply`: changes a reversible local setting only after strong evidence,
  then explains the change and offers undo.

Use `autoApply` conservatively. Never auto-apply changes involving secrets,
credentials, permissions, destructive actions, provider setup, billing-like
behavior, or hard-to-reverse choices.

## Implementation Workflow

1. Define or update the rule in `src/shared/assistive-ux/registry.ts`.
2. Keep lifecycle reads and writes in `src/shared/assistive-ux/runtime.ts` and
   `src/shared/assistive-ux/state.ts`.
3. Render the moment from the relevant feature surface.
4. Retire the moment when the user accepts it, dismisses it, sees it enough
   times, or manually changes the related setting.
5. If the moment changes a setting, call the setting's public setter and provide
   an undo path when the change is automatic.
6. Keep stored signals coarse. Do not store chat text, file paths, provider
   secrets, credentials, or message contents.

## UX Checklist

Before shipping, confirm:

- The moment appears close to the behavior it helps with.
- The copy explains the next action without selling the feature.
- The primary workflow still has visual priority.
- The moment has a clear retirement path.
- Repeated exposure is capped.
- Manual setting changes stop or reset related guidance.
- The behavior works in light and dark themes.

## Testing

Add focused tests for:

- fresh eligibility
- shown-count updates
- max-show expiration
- acceptance or dismissal retirement
- related setting-change retirement
- invalid localStorage fallback
- any feature surface that renders or applies the moment

For broad behavior, add Playwright coverage only when the flow benefits from
real app navigation or persistence across reloads.
````

## File: README.md
upstream: https://github.com/block/berd/blob/main/README.md

````text
# Berd

Berd is an open-source desktop app for working with AI agents. It is built with
Tauri 2 and React 19 and talks to the upstream Goose backend through the ACP
WebSocket served by a `goose serve` sidecar.

The repository builds a general-purpose public distribution. Organizations can
also create enterprise distributions by supplying managed provider settings,
private resources, and release infrastructure through the repository's
distribution seams without adding private material to the public source tree.

## Getting started

```bash
just setup
just dev
```

`just setup` installs pnpm dependencies, builds the vendored `@aaif/goose-sdk`,
and prepares the Goose backend pinned by `goose-backend.lock.json` in your
platform cache directory. `just dev` reuses that stamped pinned binary and fails
if the lockfile commit no longer matches the cached build.

If you already have an upstream Goose binary you want to test, set
`GOOSE_BIN=/path/to/goose` before running `just dev`; that is an explicit local
override and bypasses the managed pinned checkout.

To bump the default Goose backend, update the lockfile in a PR:

```bash
scripts/update-goose-backend-lock.sh main # or a tag/branch/sha
just goose-sync                          # fetch/build the new pinned commit
```

## Bundling and distributions

Tauri bundles the Goose backend as an external sidecar. By default, `just bundle`
stages the pinned managed Goose binary from `goose-backend.lock.json` and then
runs `pnpm tauri build`:

```bash
just bundle
```

You can stage an explicit local binary with
`GOOSE_BIN=/path/to/goose just stage-sidecar`. Staging creates
`src-tauri/binaries/goosed-<rust-host-triple>`, matching the
`"externalBin": ["binaries/goosed"]` entry in `src-tauri/tauri.conf.json`.

The public build is self-contained and does not require private package
registries or enterprise credentials. Enterprise distributors may overlay
private agents, runtime configuration, optional companion tools, update
channels, and signing or publishing infrastructure in their own private build
orchestration.

## Optional companion CLI

Berd includes a distribution seam for bundling an optional companion CLI as an
app resource. The public app does not require a private CLI package; enterprise
distributors can provide and package their own implementation while retaining
the normal Berd build and validation flow.

## Public Agent Skills

Berd publishes portable Agent Skills under [`skills/`](skills/README.md). These
can be installed independently of the Berd app and are separate from the
contributor workflows under `.agents/skills/` and the starter skills bundled
under `distro/skills/`.

The first published skill, [`buzz-handoff`](skills/buzz-handoff/SKILL.md), brings
Buzz channel or thread context into a private agent conversation and can send an
explicitly approved reply through the public Buzz CLI.

## Adding an experiment

Experiments are user-local preferences for unstable UI or workflow behavior.
Untouched experiments follow the global auto-enable setting, which defaults on
in dev builds and off in production builds. Use
`.agents/skills/experimental-features/SKILL.md` for the current workflow,
registry contract, storage rules, Tauri guardrails, and test coverage.

## Participating

Berd is built by a small team at Block, in the open. You can read the source,
build it, and fork it freely — but **we don't accept pull requests from outside
authorized repository collaborators**, and outside PRs are closed automatically.

The way to participate is to **open a well-formed issue**. A bug report we can
reproduce is worth more to us than a patch, because it's the part we can't do
ourselves. [CONTRIBUTING.md](CONTRIBUTING.md) spells out exactly what each kind
of issue needs; the [issue forms](https://github.com/block/berd/issues/new/choose)
require it.

Filing one? Hand this to your coding agent:

```
Read https://raw.githubusercontent.com/block/berd/main/CONTRIBUTING.md
and help me file a Berd issue. Interview me for anything the guide
requires that I haven't given you, and tell me if what I'm reporting
is actually two separate issues.
```

Please also review the [Code of Conduct](CODE_OF_CONDUCT.md) and
[Security Policy](SECURITY.md). Never report a security vulnerability as a
public issue.

## Useful commands

- `just check` — Biome, design-system, i18n, contract, and type checks
- `just test` — unit and component tests
- `just tauri-check` — Rust type check with sidecars disabled
- `just clippy` — Rust lint with warnings denied
- `just bundle` — stage the pinned Goose backend and run `pnpm tauri build`
````
