@kendoo.agentdesk/agentdesk 0.21.0 → 0.22.1

package/CHANGELOG.md

@@ -13,6 +13,45 @@ Internal refactors, infrastructure changes, and architectural notes are not list
 ### Added
 - `[UI]` Private-session sharing flow. When someone visits a session URL they don't own, they now see a friendly "Shh… this one's private" page with a one-click "Request access" button instead of a blank error. The session owner sees pending requests in a new header inbox and can grant viewer access for just that session or the whole project. Viewer-granted sessions show up in the teammate's sidebar tagged as a viewer.
 
+### Changed
+- `[Both]` Security improvements and hardening. No action required.
+
+## [0.22.0] — 2026-05-23
+
+### Changed
+- `[CLI]` `agentdesk team` now defaults to phased mode so per-phase model overrides actually apply. Previously the whole session ran in a single process, which meant the model configured for INTAKE was silently overridden by EXECUTION's model (or vice versa). The fix routes each phase through its own Claude process, honoring per-phase model selection. Pass `--single-process` (or `--no-phased`) to opt back into single-process behavior.
+- `[Both]` Daemon-cancelled sessions now report `status="stopped"` (terminal) instead of `"complete"`. The dashboard renders Stopped in red and surfaces Error and Canceled distinctly instead of collapsing them into a generic "Disconnected" badge.
+
+### Added
+- `[CLI]` New CLI-created tasks auto-assign to the connected user across Jira, Linear, and GitHub. No more "no assignee" tasks landing on the tracker after team kickoff.
+- `[CLI]` Orchestrator prints the resolved model per phase to stderr — useful when verifying per-phase overrides are actually taking effect.
+- `[CLI]` Stream parser detects phase markers per-line (canonical `# PHASE: <NAME>`, heading-style, and natural-language transitions) so the dashboard timeline doesn't get stuck on long-running phases.
+
+### Fixed
+- `[UI]` Daemon `heartbeat` now bumps `lastActivityAt` for the user's active sessions and revives stale ones — prevents misleading "CLI Offline" indicators during deep tool work when the CLI process is alive but Claude is busy in a long bash or edit run.
+- `[UI]` `ProjectSettingsPage` copy reflects that per-phase models always apply now (was previously hedged from the days when single-process runs ignored the override).
+
+## [0.21.0] — 2026-04-28
+
+Mobile-design wave plus session-UI polish. Versions `0.20.4` through `0.20.6` were intermediate publishes during the same wave on the same day and are superseded by `0.21.0`.
+
+### Added
+- `[UI]` Full mobile design rolled out across the dashboard, Sessions list, Run screen, composer, and chat surfaces (AD-14 through AD-25). Includes hybrid design tokens, typography aligned with claude/design, mobile chat status bar, magazine-style Sessions header, and bigger chat text on mobile.
+- `[UI]` Topbar mono breadcrumb and ⌘↵ hint on the Run Team page.
+- `[UI]` Transcript run banner and completion footer with article-width body.
+- `[UI]` Composer "›" prefix and a keyboard hint row surfacing the ⌘↵ binding.
+
+### Changed
+- `[UI]` Dashboard is now a 2-column layout (metric strip on top, team foot on bottom).
+- `[UI]` `SessionList` restyle — sharper visuals, selected-only highlighting (no more "row hover lights up every other row").
+
+### Fixed
+- `[UI]` AgentDesk wordmark font no longer regresses to the wrong typeface after the design-token rollout.
+- `[UI]` Settings form state now resets when you switch projects (was previously carrying stale fields from the prior project).
+
+### Removed
+- `[UI]` Unused `SessionStats` component.
+
 ## [0.20.3] — 2026-04-19
 
 ### Fixed

package/README.md

@@ -21,7 +21,31 @@ You can also add **custom agents** (e.g., a Security Engineer or DevOps speciali
 
 ## Getting Started
 
-### 1. Install
+The fastest way to try AgentDesk is the **Claude Code plugin**. It runs the phased loop (intake → plan → execute → review → summary) entirely inside your existing Claude Code session — no signup, no daemon, no extra install.
+
+### Quickstart — Claude Code plugin
+
+Inside any Claude Code session:
+
+```text
+/plugin marketplace add anthropics/claude-plugins-community
+/plugin install agentdesk@claude-community
+/agentdesk <your task or tracker URL>
+```
+
+Five sub-agents run in sequence with a dedicated review pass before SUMMARY produces the final report. State persists in `.agentdesk/session-memory.md`.
+
+**What the plugin covers:** the phased workflow, one task at a time, in your terminal. Zero-account, in-process.
+
+**What needs the full CLI + dashboard (below):** session history, dashboard visibility, multi-session parallelism, team coordination, tracker integration (Jira/Linear/GitHub write-back), and the long-running daemon.
+
+---
+
+### Full setup — CLI + dashboard
+
+For session history, team workflows, and tracker integration, install the AgentDesk CLI and connect it to your `agentdesk.live` account.
+
+#### 1. Install
 
 ```bash
 npm i -g @kendoo.agentdesk/agentdesk
@@ -29,7 +53,7 @@ npm i -g @kendoo.agentdesk/agentdesk
 
 Requires [Claude Code](https://claude.ai/code) and [Node.js](https://nodejs.org/) 18+.
 
-### 2. Sign in
+#### 2. Sign in
 
 ```bash
 agentdesk login
@@ -37,7 +61,7 @@ agentdesk login
 
 Opens your browser to [agentdesk.live](https://agentdesk.live) where you sign up or log in (email/password or Google). Your API key is saved automatically.
 
-### 3. Set up your project
+#### 3. Set up your project
 
 ```bash
 agentdesk init
@@ -71,7 +95,7 @@ All menus use arrow-key navigation with color highlighting; Ctrl-C exits cleanly
 
 `.agentdesk.json` is kept in sync with the server: every CLI run fetches the authoritative config and rewrites the local file as a credential-free snapshot. Credentials live only on the server, encrypted. If a local edit disagreed with the server's value, a one-line warning lists the overridden fields.
 
-### 4. Run a team session
+#### 4. Run a team session
 
 Work on an existing task:
 

package/cli/tracker-check.mjs

@@ -1,7 +1,12 @@
 // Verify tracker API permissions before starting a session
 // Checks: read tasks, create tasks, update tasks
 
-import { execSync } from "child_process";
+import { execFileSync } from "child_process";
+
+// AD-29: any external string (repo name, ref, etc.) must be regex-validated
+// before it goes into a child process. Even with execFile, an upstream caller
+// passing "../../etc" or similar can trip path semantics in the called tool.
+const GITHUB_REPO_RE = /^[A-Za-z0-9._-]+\/[A-Za-z0-9._-]+$/;
 
 const LINEAR_API = "https://api.linear.app/graphql";
 
@@ -187,18 +192,25 @@ async function checkGitHub({ repo }, creds) {
   const env = { ...process.env };
   if (creds?.GITHUB_TOKEN) env.GH_TOKEN = creds.GITHUB_TOKEN;
 
-  // Check if gh CLI is available
+  // AD-29: refuse anything that isn't a clean owner/repo before going near a
+  // subprocess. `repo` comes from server-pushed project settings, which an
+  // attacker could control via a chained takeover.
+  if (!GITHUB_REPO_RE.test(repo)) {
+    return { ok: false, errors: [`Invalid GitHub repo format: ${JSON.stringify(repo)} (expected "owner/name")`] };
+  }
+
+  // Check if gh CLI is available. execFileSync (no shell) — argv is static.
   try {
-    execSync("gh --version", { stdio: "pipe" });
+    execFileSync("gh", ["--version"], { stdio: "pipe" });
   } catch {
     return { ok: false, errors: ["GitHub CLI (gh) is not installed — install it from https://cli.github.com"] };
   }
 
   // Check auth status and grab the authenticated login
   try {
-    execSync("gh auth status", { stdio: "pipe", env });
+    execFileSync("gh", ["auth", "status"], { stdio: "pipe", env });
     try {
-      const who = execSync(`gh api user --jq '{login: .login, name: .name}'`, { stdio: "pipe", encoding: "utf-8", env }).trim();
+      const who = execFileSync("gh", ["api", "user", "--jq", "{login: .login, name: .name}"], { stdio: "pipe", encoding: "utf-8", env }).trim();
       const j = JSON.parse(who);
       identity = { name: j.name || j.login, login: j.login };
     } catch {}
@@ -206,9 +218,9 @@ async function checkGitHub({ repo }, creds) {
     return { ok: false, errors: ["GitHub CLI is not authenticated — run 'gh auth login', or paste a GITHUB_TOKEN"] };
   }
 
-  // Check repo access and permissions
+  // Check repo access and permissions — argv form, no shell, repo already validated.
   try {
-    const result = execSync(`gh api repos/${repo} --jq ".permissions"`, { stdio: "pipe", encoding: "utf-8", env });
+    const result = execFileSync("gh", ["api", `repos/${repo}`, "--jq", ".permissions"], { stdio: "pipe", encoding: "utf-8", env });
     const perms = JSON.parse(result.trim());
 
     if (!perms.pull) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kendoo.agentdesk/agentdesk",
-  "version": "0.21.0",
+  "version": "0.22.1",
   "description": "AI team orchestrator for Claude Code — run collaborative agent sessions from your terminal",
   "type": "module",
   "bin": {
@@ -22,7 +22,7 @@
     "server": "node server/index.mjs",
     "build": "vite build",
     "preview": "vite preview",
-    "test": "node --test tests/server.test.mjs tests/agents.test.mjs tests/homepage.test.mjs tests/sessionUtils.test.mjs tests/tracker-url.test.mjs tests/session-preflight.test.mjs tests/access.test.mjs",
+    "test": "node --test tests/server.test.mjs tests/agents.test.mjs tests/homepage.test.mjs tests/sessionUtils.test.mjs tests/tracker-url.test.mjs tests/session-preflight.test.mjs tests/access.test.mjs tests/project-ownership.test.mjs tests/random-hex.test.mjs",
     "lint:changelog": "node scripts/lint-changelog.mjs",
     "prepublishOnly": "node scripts/lint-changelog.mjs"
   },