@vellumai/assistant 0.4.56 → 0.4.57

package/src/config/bundled-skills/acp/TOOLS.json

@@ -3,7 +3,7 @@
   "tools": [
     {
       "name": "acp_spawn",
-      "description": "Spawn an external coding agent (e.g. Claude Code, Codex, Gemini CLI) via ACP to work on a task. The agent runs as a subprocess and streams results back. Use this when you want to delegate a coding task to an external agent that has its own tools, file editing, and terminal access. If ACP is not enabled, follow the setup instructions in SKILL.md to install claude-agent-acp and configure it. The command MUST be 'claude-agent-acp' — NEVER use 'claude', 'claude -p', or 'claude --acp'.",
+      "description": "Spawn an external coding agent (e.g. Claude Code, Codex, Gemini CLI) via ACP to work on a task. The agent runs as a subprocess and streams results back. Use this when you want to delegate a coding task to an external agent that has its own tools, file editing, and terminal access. If ACP is not enabled, follow the setup instructions in SKILL.md to install claude-agent-acp and configure it. The command MUST be 'claude-agent-acp' - NEVER use 'claude', 'claude -p', or 'claude --acp'.",
       "category": "orchestration",
       "risk": "low",
       "input_schema": {
@@ -15,13 +15,13 @@
           },
           "task": {
             "type": "string",
-            "description": "The task to give the agent — what it should accomplish"
+            "description": "The task to give the agent - what it should accomplish"
           },
           "cwd": {
             "type": "string",
             "description": "Working directory for the agent. This determines where the agent runs and where its session is stored. Set this to the project/repo root the user wants the agent to work in. Defaults to current conversation's working directory."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief explanation of why this tool is being called"
           }
@@ -33,7 +33,7 @@
     },
     {
       "name": "acp_status",
-      "description": "Get the status of a specific ACP session or list all ACP sessions. Only use this when the user explicitly asks about ACP session status — do NOT poll automatically, as you will be notified when sessions complete.",
+      "description": "Get the status of a specific ACP session or list all ACP sessions. Only use this when the user explicitly asks about ACP session status - do NOT poll automatically, as you will be notified when sessions complete.",
       "category": "orchestration",
       "risk": "low",
       "input_schema": {
@@ -43,7 +43,7 @@
             "type": "string",
             "description": "Optional ACP session ID to query. If omitted, returns all ACP sessions."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief explanation of why this tool is being called"
           }
@@ -65,7 +65,7 @@
             "type": "string",
             "description": "The ID of the ACP session to abort."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief explanation of why this tool is being called"
           }

package/src/config/bundled-skills/app-builder/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: app-builder
-description: Build interactive apps, dashboards, calculators, games, trackers, tools, landing pages, and data visualizations with HTML/CSS/JS. Use when the user says build, create, or make an app/dashboard/calculator/game.
+description: Build interactive apps, dashboards, calculators, games, trackers, tools, landing pages, and data visualizations with HTML/CSS/JS
 compatibility: "Designed for Vellum personal assistants"
 metadata:
   emoji: "🏗️"
@@ -10,9 +10,9 @@ metadata:
       - "frontend-design"
 ---
 
-You are an expert app builder and visual designer. When the user asks you to create an app, tool, or utility, you immediately design a data schema, choose a stunning visual direction, build the interface, and open it — all in one step. You don't discuss or ask for permission to be creative. You ARE the designer: you pick the colors, the layout, the atmosphere, the micro-interactions. Your apps should make users stop and say "whoa" — they should feel designed, not generated.
+You are an expert app builder and visual designer. When the user asks you to create an app, tool, or utility, you immediately design a data schema, choose a stunning visual direction, build the interface, and open it - all in one step. You don't discuss or ask for permission to be creative. You ARE the designer: you pick the colors, the layout, the atmosphere, the micro-interactions. Your apps should make users stop and say "whoa" - they should feel designed, not generated.
 
-**Every app gets its own visual identity.** A plant tracker should feel earthy and green. A finance dashboard should feel precise and navy. A fitness app should feel energetic and purple. Apps should look like they were designed by a boutique studio for that specific domain — not like generic branded tools. Think standalone premium product, not template.
+**Every app gets its own visual identity.** A plant tracker should feel earthy and green. A finance dashboard should feel precise and navy. A fitness app should feel energetic and purple. Apps should look like they were designed by a boutique studio for that specific domain - not like generic branded tools. Think standalone premium product, not template.
 
 **Your default behavior:** Build immediately. The user types "build me a habit tracker" and you deliver a complete, polished app with a domain-matched color palette, atmospheric background, and thoughtful interactions. Don't ask what colors they want. Don't show wireframes. Just build something stunning and let them refine from there.
 
@@ -24,7 +24,7 @@ You are an expert app builder and visual designer. When the user asks you to cre
 
 **Default: just build.** When a user says "build me a habit tracker," don't ask what colors they want or how many fields to include. Immediately:
 
-1. Envision the ideal version of this app — what would make someone excited to use it?
+1. Envision the ideal version of this app - what would make someone excited to use it?
 2. Pick a distinctive visual direction following the `frontend-design` skill
 3. Design a clean data schema
 4. Build the complete, polished app with animations, interactions, and empty states
@@ -42,15 +42,15 @@ You are an expert app builder and visual designer. When the user asks you to cre
 
 <!-- feature:app-builder-multifile:alt:end -->
 
-**Only ask questions when the request is genuinely ambiguous** — e.g., "build me an app" with no indication of what kind. Even then, prefer building something impressive based on context clues over asking a battery of questions.
+**Only ask questions when the request is genuinely ambiguous** - e.g., "build me an app" with no indication of what kind. Even then, prefer building something impressive based on context clues over asking a battery of questions.
 
-**When in doubt, build something impressive** and let the user refine with `app_update`. The first impression matters most — a beautiful app with the wrong shade of blue is easy to fix. A correct but ugly app is hard to come back from.
+**When in doubt, build something impressive** and let the user refine with `app_update`. The first impression matters most - a beautiful app with the wrong shade of blue is easy to fix. A correct but ugly app is hard to come back from.
 
 **There are no "quick" builds.** Every app, regardless of complexity, gets the full design treatment. A 3-field form and a 20-section dashboard get the same design care. The only difference is scope, not quality.
 
 ### 2. Design the Data Schema
 
-Create a JSON Schema that defines the structure of a single record. Every record automatically gets `id`, `appId`, `createdAt`, and `updatedAt` — you only define user-facing fields.
+Create a JSON Schema that defines the structure of a single record. Every record automatically gets `id`, `appId`, `createdAt`, and `updatedAt` - you only define user-facing fields.
 
 Schema guidelines:
 
@@ -58,7 +58,7 @@ Schema guidelines:
 - Define `properties` for each field
 - Supported types: `string`, `number`, `boolean`
 - Add a `required` array for mandatory fields
-- Keep schemas reasonably flat — encode complex nested data as JSON strings when needed
+- Keep schemas reasonably flat - encode complex nested data as JSON strings when needed
 
 Example schema for a project tracker:
 
@@ -96,8 +96,8 @@ Build apps as multi-file TSX projects. You get component reuse, TypeScript type-
 
 ```
 src/
-  index.html          # Entry HTML — minimal shell, loads compiled bundle
-  main.tsx             # App entry — renders root component into #app
+  index.html          # Entry HTML - minimal shell, loads compiled bundle
+  main.tsx             # App entry - renders root component into #app
   components/          # Preact functional components
     Header.tsx
     RecordList.tsx
@@ -139,7 +139,7 @@ export const Header: FunctionComponent<Props> = ({ title, count }) => {
 
 **CSS:** Import CSS files directly in TSX (`import './styles.css'`). You can also use inline styles via the `style` attribute on JSX elements.
 
-**Data bridge:** The same `window.vellum.data` API works in TSX components — call it from `useEffect` hooks or event handlers:
+**Data bridge:** The same `window.vellum.data` API works in TSX components - call it from `useEffect` hooks or event handlers:
 
 ```tsx
 const [records, setRecords] = useState<Record[]>([]);
@@ -151,9 +151,9 @@ useEffect(() => {
 
 **File workflow:** Use `app_file_write` for each source file. Each write automatically triggers a recompile of the project.
 
-**Allowed third-party packages:** `date-fns`, `chart.js`, `lodash-es`, `zod`, `clsx`, `lucide`. Import them directly — esbuild resolves them at build time. No CDN imports. Note: `lucide` is the vanilla JS icon library (not `lucide-react`). Use its `createElement` or `createIcons` API, or manually inline SVG — do not import JSX icon components.
+**Allowed third-party packages:** `date-fns`, `chart.js`, `lodash-es`, `zod`, `clsx`, `lucide`. Import them directly - esbuild resolves them at build time. No CDN imports. Note: `lucide` is the vanilla JS icon library (not `lucide-react`). Use its `createElement` or `createIcons` API, or manually inline SVG - do not import JSX icon components.
 
-**Example — creating a multi-file project:**
+**Example - creating a multi-file project:**
 
 ```
 app_file_write(app_id, "src/index.html", `<!DOCTYPE html>
@@ -209,12 +209,12 @@ app_file_write(app_id, "src/styles.css", `.app { padding: var(--v-spacing-lg); }
 
 **Technical constraints (multi-file):**
 
-- No CDN imports — use esbuild-resolved packages from the allowlist above
-- Preact for UI (not React) — `import { render } from 'preact'`
+- No CDN imports - use esbuild-resolved packages from the allowlist above
+- Preact for UI (not React) - `import { render } from 'preact'`
 - TypeScript encouraged for all `.tsx`/`.ts` files
-- No external fonts, images, or resources — use system fonts and CSS/SVG for visuals
+- No external fonts, images, or resources - use system fonts and CSS/SVG for visuals
 - Design for 400-600px width with graceful resizing
-- The WebView blocks all navigation — links and form `action` attributes won't work
+- The WebView blocks all navigation - links and form `action` attributes won't work
 <!-- feature:app-builder-multifile:end -->
 
 <!-- feature:app-builder-multifile:alt -->
@@ -225,11 +225,11 @@ Write a complete, self-contained HTML document.
 
 **Technical constraints (single-file):**
 
-- Single HTML string — no external files, CDNs, or imports
+- Single HTML string - no external files, CDNs, or imports
 - All CSS in `<style>` in `<head>`, all JavaScript in `<script>` before `</body>`
-- No external fonts, images, or resources — use system fonts and CSS/SVG for visuals
+- No external fonts, images, or resources - use system fonts and CSS/SVG for visuals
 - Design for 400-600px width with graceful resizing
-- The WebView blocks all navigation — links and form `action` attributes won't work
+- The WebView blocks all navigation - links and form `action` attributes won't work
 
 <!-- feature:app-builder-multifile:alt:end -->
 
@@ -237,7 +237,7 @@ Write a complete, self-contained HTML document.
 
 A design system CSS is auto-injected inside a `@layer`, so your styles always take priority. It provides element defaults and automatic light/dark mode switching via `prefers-color-scheme`.
 
-**Use `--v-*` variables and `.v-*` classes** — they handle light/dark mode automatically. No manual dark mode CSS needed.
+**Use `--v-*` variables and `.v-*` classes** - they handle light/dark mode automatically. No manual dark mode CSS needed.
 
 Available design tokens:
 
@@ -269,7 +269,7 @@ window.addEventListener("vellum-theme-change", (e) => {
 
 #### Widget component library
 
-A CSS/JS widget library is auto-injected alongside the design system. Use these for standard UI patterns — skip them when custom HTML serves the user better.
+A CSS/JS widget library is auto-injected alongside the design system. Use these for standard UI patterns - skip them when custom HTML serves the user better.
 
 **Layout widgets** (class names, infer HTML structure):
 
@@ -287,7 +287,7 @@ A CSS/JS widget library is auto-injected alongside the design system. Use these
 | `.v-progress-bar` / `.v-progress-track` / `.v-progress-fill` | Horizontal progress                                            |
 | `.v-status-badge`                                            | Colored pill with dot (`.success`/`.error`/`.warning`/`.info`) |
 | `.v-stat-row` / `.v-stat`                                    | Horizontal label-value pairs                                   |
-| `.v-toast`                                                   | Notification banner — prefer `vellum.widgets.toast()`          |
+| `.v-toast`                                                   | Notification banner - prefer `vellum.widgets.toast()`          |
 | `.v-avatar-row`                                              | Contact/team display                                           |
 | `.v-tag-group`                                               | Wrapping tag row                                               |
 
@@ -396,26 +396,26 @@ vellum.widgets.slideshow("deck", {
 
 #### When to use widgets vs custom HTML
 
-- **Use widgets** for standard patterns — tables, metrics, timelines, notifications, presentations
-- **Use custom HTML** for novel or creative UIs — games, art tools, unique dashboards
-- **Mix freely** — widgets compose well together and with custom elements
+- **Use widgets** for standard patterns - tables, metrics, timelines, notifications, presentations
+- **Use custom HTML** for novel or creative UIs - games, art tools, unique dashboards
+- **Mix freely** - widgets compose well together and with custom elements
 - **ALWAYS use `vellum.widgets.*` chart functions** instead of hand-coding SVG/CSS charts. They handle overflow clipping, bounds, scaling, and dark mode. Hand-coded charts break layouts.
 
 #### Data bridge API
 
 The HTML interface can read and write records via `window.vellum.data`. All methods return Promises.
 
-- `window.vellum.data.query()` — Returns all records: `{ id, appId, data, createdAt, updatedAt }[]`
-- `window.vellum.data.create(data)` — Creates a record. Returns the created record.
-- `window.vellum.data.update(recordId, data)` — Updates a record by ID. Returns updated record.
-- `window.vellum.data.delete(recordId)` — Deletes a record by ID. Returns void.
+- `window.vellum.data.query()` - Returns all records: `{ id, appId, data, createdAt, updatedAt }[]`
+- `window.vellum.data.create(data)` - Creates a record. Returns the created record.
+- `window.vellum.data.update(recordId, data)` - Updates a record by ID. Returns updated record.
+- `window.vellum.data.delete(recordId)` - Deletes a record by ID. Returns void.
 
 Important:
 
 - Call `query()` on page load to populate initial state
 - User fields live in `record.data` (e.g., `record.data.title`)
 - Record IDs are UUID strings
-- All operations are async — use `async/await`
+- All operations are async - use `async/await`
 - Wrap all calls in `try/catch`
 
 #### Client-side state management
@@ -490,11 +490,11 @@ Call `app_create` with:
 - `name`: Short descriptive name
 - `description`: One-sentence summary
 - `schema_json`: JSON schema as string
-- `html`: (optional) Complete HTML document as string for `index.html`. If omitted, a minimal scaffold is created — you can then write `index.html` and other files via `app_file_write`.
+- `html`: (optional) Complete HTML document as string for `index.html`. If omitted, a minimal scaffold is created - you can then write `index.html` and other files via `app_file_write`.
 - `auto_open`: (optional, defaults to `true`) Shows an inline preview card in chat
-- `preview`: Always include — `title` (required), `subtitle`, `description`, `icon` (image URL preferred, emoji fallback), `metrics` (up to 3 key-value pills)
+- `preview`: Always include - `title` (required), `subtitle`, `description`, `icon` (image URL preferred, emoji fallback), `metrics` (up to 3 key-value pills)
 
-The app is NOT opened in a workspace panel automatically — users open it via the 'Open App' button on the inline card.
+The app is NOT opened in a workspace panel automatically - users open it via the 'Open App' button on the inline card.
 
 <!-- feature:app-builder-multifile:start -->
 ### 5. Handle Iteration
@@ -505,9 +505,9 @@ The app is NOT opened in a workspace panel automatically — users open it via t
 
 When the user requests changes, prefer **`app_file_edit`** over rewriting the entire file.
 
-- **`app_file_edit`** — preferred for targeted changes (styles, bugs, features). Provide `app_id`, `path`, `old_string`, `new_string`.
-- **`app_file_write`** — use when creating new files or when changes are so extensive a full rewrite is cleaner.
-- **`app_update`** — metadata only: `name`, `description`, `schema_json`. Not for code changes.
+- **`app_file_edit`** - preferred for targeted changes (styles, bugs, features). Provide `app_id`, `path`, `old_string`, `new_string`.
+- **`app_file_write`** - use when creating new files or when changes are so extensive a full rewrite is cleaner.
+- **`app_update`** - metadata only: `name`, `description`, `schema_json`. Not for code changes.
 - Always include a **`status`** parameter describing what you're doing.
 
 Apps can have multiple files (`styles.css`, `app.js`, etc.). Link from `index.html` with standard tags.
@@ -528,10 +528,10 @@ Slides are a different domain from apps. Skip app-specific patterns (contextual
 
 **Key principles:**
 
-- One idea per slide — understood in 3 seconds
-- Layout variety — 3+ different types per deck, never consecutive same-type
+- One idea per slide - understood in 3 seconds
+- Layout variety - 3+ different types per deck, never consecutive same-type
 - 8 layout types: Title, Stats, Bullets, Quote, Comparison, Timeline, Visual/Immersive, Closing/CTA
-- Bold backgrounds — dark, gradient, or strongly tinted
+- Bold backgrounds - dark, gradient, or strongly tinted
 - Max 6 bullets per slide, max 3 sentences body text
 - Never go below 15px for any visible text
 
@@ -540,7 +540,7 @@ Init with `vellum.widgets.slideshow()`. Use `.v-slide` with helpers: `.v-slide-l
 ## Error Handling
 
 - All `window.vellum.data` calls must be wrapped in `try/catch` with user-friendly feedback.
-- Never let a failed operation silently pass — always show a toast or inline error.
+- Never let a failed operation silently pass - always show a toast or inline error.
 - If the page loads with no data, show a designed empty state (`.v-empty-state`).
 - For forms, show validation errors inline next to the relevant field.
 
@@ -550,7 +550,7 @@ When the user wants to triage or bulk-act on items, generate an interactive UI w
 
 1. Fetch data with relevant tools
 2. Render a `dynamic_page` with selectable items and action buttons
-3. User selects + clicks action — UI sends `surfaceAction` with action ID and selected IDs
+3. User selects + clicks action - UI sends `surfaceAction` with action ID and selected IDs
 4. Execute tools, update UI with `ui_update`, show feedback via `widgets.toast()`
 5. Use `window.vellum.confirm()` for destructive actions
 

package/src/config/bundled-skills/app-builder/TOOLS.json

@@ -75,7 +75,7 @@
             },
             "required": ["title"]
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -93,7 +93,7 @@
       "input_schema": {
         "type": "object",
         "properties": {
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -114,7 +114,7 @@
             "type": "string",
             "description": "The ID of the app to query records for"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -159,7 +159,7 @@
               "type": "string"
             }
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -181,7 +181,7 @@
             "type": "string",
             "description": "The ID of the app to delete"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -203,7 +203,7 @@
             "type": "string",
             "description": "The ID of the app"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -237,7 +237,7 @@
             "type": "number",
             "description": "Number of lines to return (default: all)"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -279,7 +279,7 @@
             "type": "string",
             "description": "Optional short human-readable progress message shown to the user (e.g. 'adding dark mode styles')"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -313,7 +313,7 @@
             "type": "string",
             "description": "Optional short human-readable progress message shown to the user (e.g. 'adding dark mode styles')"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -339,7 +339,7 @@
             "type": "string",
             "description": "Optional description to guide icon generation (e.g. 'a blue calendar with a checkmark'). If omitted, the app name and description are used."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }

package/src/config/bundled-skills/browser/SKILL.md

@@ -13,24 +13,24 @@ metadata:
 
 Use this skill to browse the web. After loading this skill, the following browser tools become available:
 
-- `browser_navigate` — Navigate to a URL
-- `browser_snapshot` — List interactive elements on the current page
-- `browser_screenshot` — Take a visual screenshot
-- `browser_close` — Close the browser page
-- `browser_click` — Click an element
-- `browser_type` — Type text into an input
-- `browser_press_key` — Press a keyboard key
-- `browser_scroll` — Scroll the page or a specific element
-- `browser_select_option` — Select an option from a native `<select>` element
-- `browser_hover` — Hover over an element to reveal menus/tooltips
-- `browser_wait_for` — Wait for a condition
-- `browser_extract` — Extract page text content
-- `browser_wait_for_download` — Wait for a file download to complete
-- `browser_fill_credential` — Fill a stored credential into a form field
+- `browser_navigate` - Navigate to a URL
+- `browser_snapshot` - List interactive elements on the current page
+- `browser_screenshot` - Take a visual screenshot
+- `browser_close` - Close the browser page
+- `browser_click` - Click an element
+- `browser_type` - Type text into an input
+- `browser_press_key` - Press a keyboard key
+- `browser_scroll` - Scroll the page or a specific element
+- `browser_select_option` - Select an option from a native `<select>` element
+- `browser_hover` - Hover over an element to reveal menus/tooltips
+- `browser_wait_for` - Wait for a condition
+- `browser_extract` - Extract page text content
+- `browser_wait_for_download` - Wait for a file download to complete
+- `browser_fill_credential` - Fill a stored credential into a form field
 
 ## Capabilities
 
-This browser runs **full Chromium with JavaScript enabled**. It can handle SPAs, React/Vue/Angular apps, dynamic content, date pickers, booking systems, reservation flows, and any JavaScript-heavy interactive site. Never tell the user you "can't handle interactive JavaScript" — you can.
+This browser runs **full Chromium with JavaScript enabled**. It can handle SPAs, React/Vue/Angular apps, dynamic content, date pickers, booking systems, reservation flows, and any JavaScript-heavy interactive site. Never tell the user you "can't handle interactive JavaScript" - you can.
 
 ## Typical Workflow
 

package/src/config/bundled-skills/browser/TOOLS.json

@@ -17,7 +17,7 @@
             "type": "boolean",
             "description": "If true, allows navigation to localhost/private-network hosts. Disabled by default for SSRF safety."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -35,7 +35,7 @@
       "input_schema": {
         "type": "object",
         "properties": {
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -56,7 +56,7 @@
             "type": "boolean",
             "description": "Capture the full scrollable page instead of just the viewport."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -77,7 +77,7 @@
             "type": "boolean",
             "description": "If true, close all browser pages and the browser context. Default: false (close only the current conversation page)."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -102,7 +102,7 @@
             "type": "string",
             "description": "A CSS selector to target. Used as fallback when element_id is not available."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -139,7 +139,7 @@
             "type": "boolean",
             "description": "If true, press Enter after typing the text."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -169,7 +169,7 @@
             "type": "string",
             "description": "Optional CSS selector to target."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -204,7 +204,7 @@
             "type": "string",
             "description": "Optional CSS selector of element to scroll within."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -242,7 +242,7 @@
             "type": "number",
             "description": "The zero-based index of the <option> to select."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -267,7 +267,7 @@
             "type": "string",
             "description": "A CSS selector to target. Used as fallback when element_id is not available."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -300,7 +300,7 @@
             "type": "number",
             "description": "Maximum wait time in milliseconds (default and max: 30000)."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -321,7 +321,7 @@
             "type": "boolean",
             "description": "If true, include a list of links found on the page (up to 200)."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -342,7 +342,7 @@
             "type": "number",
             "description": "Maximum wait time in milliseconds (default: 30000, max: 120000)."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -379,7 +379,7 @@
             "type": "boolean",
             "description": "Press Enter after filling"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }

package/src/config/bundled-skills/chatgpt-import/SKILL.md

@@ -16,12 +16,12 @@ When a user wants to import their ChatGPT conversations:
 
 1. **Tell the user how to export.** They need to go to ChatGPT → Settings → Data controls → Export data. ChatGPT will email them a ZIP file.
 2. **Ask the user to upload the ZIP file.** Use whatever file upload tools or skills are available to receive the ZIP.
-3. **Run the import.** Once you have the ZIP file path, call `chatgpt_import` with the file path. Report the results — how many conversations and messages were imported, and any skipped duplicates.
+3. **Run the import.** Once you have the ZIP file path, call `chatgpt_import` with the file path. Report the results - how many conversations and messages were imported, and any skipped duplicates.
 
 ## Notes
 
 - Only ZIP files are accepted (the full export archive from ChatGPT).
-- Conversations are deduplicated — re-importing the same file will skip already-imported conversations.
+- Conversations are deduplicated - re-importing the same file will skip already-imported conversations.
 - Only user and assistant messages are imported (system prompts and tool calls are filtered out).
 - Original timestamps from ChatGPT are preserved.
 - Imported conversations are automatically indexed for memory search.

package/src/config/bundled-skills/chatgpt-import/TOOLS.json

@@ -13,7 +13,7 @@
             "type": "string",
             "description": "Absolute path to the ChatGPT export ZIP file"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }

package/src/config/bundled-skills/chatgpt-import/tools/chatgpt-import.ts

@@ -175,7 +175,7 @@ export async function run(
           memoryConfig,
         );
       } catch (err) {
-        // Indexing failure is non-fatal — the message is already persisted,
+        // Indexing failure is non-fatal - the message is already persisted,
         // and failing here would abort the loop before conversationKeys is
         // written, causing duplicate imports on retry.
         log.warn(

package/src/config/bundled-skills/claude-code/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: claude-code
-description: Delegate coding tasks to Claude Code, an AI-powered coding agent
+description: Delegate coding tasks to Claude Code
 compatibility: "Designed for Vellum personal assistants"
 metadata:
   emoji: "💻"
@@ -45,9 +45,9 @@ Do NOT delegate when:
 
 Claude Code supports scoped worker profiles that restrict tool access:
 
-- **general** (default) — Full access to all tools.
-- **researcher** — Read-only access. Can search, read files, and browse the web but cannot write or execute commands.
-- **coder** — Full read/write/execute access optimized for implementation tasks.
-- **reviewer** — Read-only access tailored for code review, with emphasis on analysis and feedback.
+- **general** (default) - Full access to all tools.
+- **researcher** - Read-only access. Can search, read files, and browse the web but cannot write or execute commands.
+- **coder** - Full read/write/execute access optimized for implementation tasks.
+- **reviewer** - Read-only access tailored for code review, with emphasis on analysis and feedback.
 
 Select the profile that best matches the task to enforce least-privilege access.

package/src/config/bundled-skills/computer-use/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: computer-use
-description: Computer-use action tools for controlling the macOS desktop via accessibility and screen capture.
+description: Control the macOS desktop
 compatibility: "Designed for Vellum personal assistants"
 metadata:
   emoji: "🖥️"
@@ -13,5 +13,5 @@ the macOS desktop. CU tools run through the main agent loop via HostCuProxy.
 
 The skill is internally preactivated for conversations with a connected desktop client.
 
-Tools in this skill are proxy tools — execution is forwarded to the connected
+Tools in this skill are proxy tools - execution is forwarded to the connected
 macOS client, never handled locally by the assistant.