projecta-rrr 1.23.5 → 1.24.0

package/AGENTS.md

@@ -31,6 +31,7 @@ The RRR loop follows five steps:
 | ------- | ----------- |
 | `$rrr-add-phase` | Add phase to end of current milestone in roadmap |
 | `$rrr-add-todo` | Capture idea or task as todo from current conversation context |
+| `$rrr-assign-phases` | Assign roadmap phases to GitHub teams for RRR team mode |
 | `$rrr-audit-milestone` | Audit milestone completion against original intent before archiving |
 | `$rrr-bootstrap-nextjs` | Scaffold a Next.js App Router MVP with TypeScript, Tailwind, shadcn/ui, Vitest, and Playwright |
 | `$rrr-brownfield-audit` | Audit brownfield repos for scattered planning docs before RRR initialization |
@@ -39,11 +40,13 @@ The RRR loop follows five steps:
 | `$rrr-check-upstream` | Run check-upstream workflow |
 | `$rrr-check-version` | Check version consistency across package.json, CHANGELOG, and planning docs |
 | `$rrr-complete-milestone` | Archive completed milestone and prepare for next version |
+| `$rrr-coordinate-merge` | Generate a GitHub-native coordinator merge readiness report |
 | `$rrr-create-roadmap` | Create roadmap with phases for the project |
 | `$rrr-debug` | Systematic debugging with persistent state across context resets |
 | `$rrr-define-requirements` | Define what "done" looks like with checkable requirements |
 | `$rrr-discuss-milestone` | Gather context for next milestone through adaptive questioning |
 | `$rrr-discuss-phase` | Gather phase context through adaptive questioning before planning |
+| `$rrr-dispatch-team` | Prepare a GitHub team workstream for execution on a branch and Sprite |
 | `$rrr-doctor` | Diagnose and fix duplicate RRR installs - non-destructive with rollback |
 | `$rrr-execute-phase` | Execute all plans in a phase with wave-based parallelization |
 | `$rrr-execute-plan` | Execute a PLAN.md file |
@@ -51,6 +54,7 @@ The RRR loop follows five steps:
 | `$rrr-help` | Show available RRR commands and usage guide |
 | `$rrr-insert-phase` | Insert urgent work as decimal phase (e.g., 72.1) between existing phases |
 | `$rrr-install-skill` | Install a skill from GitHub or marketplace |
+| `$rrr-integration-report` | Read the current RRR team-mode integration report |
 | `$rrr-list-phase-assumptions` | Surface Claude's assumptions about a phase approach before planning |
 | `$rrr-list-skills` | List all available skills (vendored and community) |
 | `$rrr-map-codebase` | Analyze codebase with parallel mapper agents to produce .planning/codebase/ documents |
@@ -72,6 +76,9 @@ The RRR loop follows five steps:
 | `$rrr-savings` | Show token-savings report (session + lifetime) with tier distribution and Opus-rate dashboard |
 | `$rrr-search-skills` | Search for skills on skillsmp.com marketplace |
 | `$rrr-ship` | Create PR from verified work with planning context, or generate SHIP-READY.md |
+| `$rrr-submit-phase` | Submit a team-owned phase for coordinator integration review |
+| `$rrr-team-init` | Initialize GitHub-native RRR team mode for a repo pilot |
+| `$rrr-team-status` | Show RRR team-mode phase assignments and submission readiness |
 | `$rrr-think` | Apply structured thinking models for architecture decisions, risk assessment, scope cuts, and root cause analysis |
 | `$rrr-update` | Update RRR to latest version with idempotent install + duplicate prevention |
 | `$rrr-verify-work` | Validate built features through audit (default) or interactive UAT |
@@ -84,4 +91,4 @@ The RRR loop follows five steps:
 - Use `$rrr-help` for full skill catalogue
 - Trigger phrases are case-sensitive: `$rrr-plan-phase` not `$rrr-PlanPhase`
 
-<!-- generated: 2026-04-20T21:10:01.139Z | source: commands/rrr/*.md | count: 47 skills -->
+<!-- generated: 2026-05-12T18:36:28.466Z | source: commands/rrr/*.md | count: 54 skills -->

package/CHANGELOG.md

@@ -4,6 +4,20 @@ All notable changes to RRR will be documented in this file.
 
 Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
+## [1.24.0] - 2026-05-12
+
+### Added
+- **Team Mode Pilot** — added `$rrr-team-init`, `$rrr-assign-phases`, `$rrr-dispatch-team`, `$rrr-team-status`, `$rrr-submit-phase`, `$rrr-coordinate-merge`, and `$rrr-integration-report` for GitHub-native team workstreams.
+- **Workstream state isolation** — teams now write to `.planning/workstreams/<team>/STATE.md`, `ASSIGNMENT.md`, and `SUBMISSION.json` so multiple developers can work on the same milestone without fighting over root `.planning/STATE.md`.
+- **GitHub + Sprites SOP** — added `docs/team-mode-github-sprites.md` and `rrr/references/team-mode.md` covering phase assignment, branch naming, Sprite sandbox names, CODEOWNERS handoff, and coordinator merge review.
+
+### Changed
+- **Browser UAT default** — RRR now prefers GSD Browser for interactive browser UAT before falling back to Agent Browser, Claude Chrome, or Playwright.
+- **Upstream monitoring** — watcher coverage now includes `gsd-build/get-shit-done`, `gsd-build/gsd-browser`, `obra/superpowers`, and `vercel-labs/agent-browser`.
+
+### Fixed
+- **Package hygiene** — generated `.rrr` semantic-search state under `rrr/lib` is excluded from npm packages.
+
 ## [1.23.5] - 2026-04-30
 
 ### Fixed

package/README.md

@@ -107,6 +107,27 @@ That's it. Verify with `/rrr:help` inside your Claude Code interface.
 
 **MVP Definition of Done at Projecta:** local demo runs + tests pass.
 
+### Team Mode Pilot
+
+RRR can pilot GitHub-native team execution where the coordinator assigns phases to GitHub teams and each team executes in an isolated branch/Sprite.
+
+```bash
+/rrr:team-init --github-org PA-MATRIX --milestone v1.24 --runner sprites
+/rrr:assign-phases --team team-runtime --phases 92,95
+/rrr:dispatch-team --team team-runtime
+/rrr:team-status
+```
+
+Developers run existing execution commands with `--team <team>` so root `.planning/STATE.md` stays coordinator-owned:
+
+```bash
+/rrr:execute-phase 92 --team team-runtime
+/rrr:verify-work 92 --team team-runtime
+/rrr:submit-phase 92 --team team-runtime --ready
+```
+
+See `docs/team-mode-github-sprites.md` for the full pilot SOP.
+
 ### Pushpa Mode (Overnight Autopilot)
 
 Run phases unattended while you sleep. **Best run in a normal terminal** (outside Claude Code):
@@ -395,22 +416,23 @@ RRR uses a deterministic verification ladder with **mandatory two-step verificat
 
 | frontend_impact | unit_tests | playwright | browser_uat |
 |-----------------|------------|------------|-------------|
-| true | Yes | Yes | Yes (agent-browse or claude --chrome) |
+| true | Yes | Yes | Yes (gsd-browser default; fallbacks available) |
 | false | Yes | No | No |
 
 **Two-Step Visual Verification (frontend_impact: true):**
 1. **Step 1: Playwright** — Automated headless E2E tests
-2. **Step 2: Browser UAT** — Natural language browser automation via agent-browse (preferred) or `claude --chrome` (fallback)
+2. **Step 2: Browser UAT** — Browser automation via GSD Browser by default, with agent-browse or `claude --chrome` as fallbacks
 
 **Browser UAT Tool Selection:**
-- **ANTHROPIC_API_KEY set + agent-browse plugin:** Uses agent-browse (natural language browser automation via Stagehand)
-- **ANTHROPIC_API_KEY not set + interactive:** Falls back to `claude --chrome`
-- **Pushpa/CI without ANTHROPIC_API_KEY:** Skipped (no automated browser available)
+- **gsd-browser installed:** Uses GSD Browser for browser control, screenshots, snapshots, and assertions
+- **gsd-browser missing + ANTHROPIC_API_KEY set + agent-browse plugin:** Falls back to agent-browse
+- **gsd-browser missing + interactive:** Falls back to `claude --chrome`
+- **Pushpa/CI without gsd-browser or ANTHROPIC_API_KEY:** Skipped
 
-**Installing agent-browse plugin (in Claude Code):**
+**GSD Browser setup:**
 ```bash
-/plugin marketplace add browserbase/agent-browse
-/plugin install agent-browse@browserbase
+# Ensure gsd-browser is on PATH, or point RRR at the binary:
+export GSD_BROWSER_BIN="$HOME/.gsd-browser/bin/gsd-browser"
 ```
 
 Both steps run even when Playwright passes — we want visual/UI confirmation.
@@ -495,14 +517,15 @@ FRONTEND_IMPACT=true bash scripts/browser-uat.sh  # Run browser UAT step
 ```
 
 **Browser UAT Tool Selection:**
-- **agent-browse** (preferred): Requires `ANTHROPIC_API_KEY` and [browserbase/agent-browse](https://github.com/browserbase/agent-browse) plugin. Uses Stagehand for natural language browser automation.
-- **claude --chrome** (fallback): Interactive mode only. Falls back when no API key set.
-- **Skipped**: In Pushpa/CI without `ANTHROPIC_API_KEY` (no automated browser available).
+- **gsd-browser** (default): Uses local CDP browser automation with screenshots, snapshots, console logs, and assertions.
+- **agent-browse** (fallback): Requires `ANTHROPIC_API_KEY` and [browserbase/agent-browse](https://github.com/browserbase/agent-browse) plugin.
+- **claude --chrome** (interactive fallback): Used only when gsd-browser and agent-browse are unavailable.
+- **Skipped**: In Pushpa/CI without gsd-browser or `ANTHROPIC_API_KEY`.
 
-**Installing agent-browse (in Claude Code):**
+**GSD Browser setup:**
 ```bash
-/plugin marketplace add browserbase/agent-browse
-/plugin install agent-browse@browserbase
+# Ensure gsd-browser is on PATH, or point RRR at the binary:
+export GSD_BROWSER_BIN="$HOME/.gsd-browser/bin/gsd-browser"
 ```
 
 **Interactive verification (UI_AFFECTING plans only):**
@@ -515,7 +538,7 @@ FRONTEND_IMPACT=true bash scripts/browser-uat.sh  # Browser UAT check
 **Integration:**
 - **Per-plan:** Runs automatically after `/rrr:execute-plan` when `frontend_impact: true` (Playwright + Browser UAT)
 - **Per-phase:** Runs automatically after `/rrr:execute-phase` (aggregates all plans)
-- **Pushpa Mode:** Headless Playwright + agent-browse (if `ANTHROPIC_API_KEY` set); no interactive fallback
+- **Pushpa Mode:** Headless Playwright + gsd-browser; agent-browse fallback if `ANTHROPIC_API_KEY` set; no interactive fallback
 - **Backend-only plans** (`frontend_impact: false`): Unit tests only, visual steps skipped
 - Results logged to `.planning/VISUAL_PROOF.md` (append-only)
 

package/bin/install.js

@@ -1746,6 +1746,7 @@ function install(isGlobal) {
     'bootstrap-external-skills.js',
     'gen-agents-md.js',
     'jarvis.sh',
+    'rrr-team-mode.js',
     'verify-milestone.js',
   ];
   for (const scriptName of additionalRuntimeScripts) {

package/commands/rrr/assign-phases.md

@@ -0,0 +1,57 @@
+---
+name: rrr:assign-phases
+description: Assign roadmap phases to GitHub teams for RRR team mode
+argument-hint: "--team <team> --phases <n,n> | interactive"
+allowed-tools:
+  - Read
+  - Write
+  - Edit
+  - Bash
+  - AskUserQuestion
+---
+
+<objective>
+Assign phases to GitHub teams as the primary team-mode work split.
+
+This command avoids hardcoded repo path ownership. The coordinator assigns outcome-oriented phases to teams; GitHub PRs, CODEOWNERS, and checks validate the actual files changed later.
+</objective>
+
+<process>
+
+1. **Load context**
+   - Read `.planning/STATE.md`.
+   - Read `.planning/ROADMAP.md`.
+   - Read `.planning/team-mode.json`.
+   - If team mode is missing, run `/rrr:team-init` first.
+
+2. **Assign**
+   If explicit args are provided:
+
+   ```bash
+   node ~/.claude/rrr/scripts/rrr-team-mode.js assign --team <team> --phases <phase-list>
+   ```
+
+   Development source repo fallback:
+
+   ```bash
+   node scripts/rrr-team-mode.js assign --team <team> --phases <phase-list>
+   ```
+
+   If no explicit args:
+   - Present phases from ROADMAP.md.
+   - Ask coordinator to assign each phase to one of the configured GitHub teams.
+   - Run the helper once per team with comma-separated phase numbers.
+
+3. **Output**
+   Show updated:
+   - `.planning/team-mode.json`
+   - `.planning/WORKSTREAMS.md`
+   - `.planning/workstreams/<team>/ASSIGNMENT.md`
+
+</process>
+
+<constraints>
+- Coordinator-only command.
+- Do not split by file paths during assignment.
+- Do not create GitHub PRs or Sprites here; dispatch does that later.
+</constraints>

package/commands/rrr/check-browser-uat.md

@@ -10,7 +10,22 @@ Check browser UAT prerequisites and report what tool will be used for visual ver
 <behavior>
 Run these checks in order:
 
-## Check 1: API Key
+## Check 1: GSD Browser
+
+```bash
+if command -v gsd-browser >/dev/null 2>&1; then
+  echo "✓ gsd-browser found: $(command -v gsd-browser)"
+  gsd_browser_present=true
+elif [[ -x "$HOME/.gsd-browser/bin/gsd-browser" ]]; then
+  echo "✓ gsd-browser found: $HOME/.gsd-browser/bin/gsd-browser"
+  gsd_browser_present=true
+else
+  echo "✗ gsd-browser not found"
+  gsd_browser_present=false
+fi
+```
+
+## Check 2: API Key (Fallback)
 
 ```bash
 if [[ -n "$ANTHROPIC_API_KEY" ]]; then
@@ -22,7 +37,7 @@ else
 fi
 ```
 
-## Check 2: agent-browse Plugin
+## Check 3: agent-browse Plugin (Fallback)
 
 Check if agent-browse MCP tools are available:
 - Look for `mcp__agent-browse__*` tools in available tools
@@ -37,7 +52,7 @@ else:
   plugin_installed=false
 ```
 
-## Check 3: Dev Server (Optional)
+## Check 4: Dev Server (Optional)
 
 Try to reach localhost:3000 (or BASE_URL if set):
 
@@ -51,7 +66,7 @@ else
 fi
 ```
 
-## Check 4: Environment Detection
+## Check 5: Environment Detection
 
 ```bash
 if [[ -n "$PUSHPA_MODE" ]] || [[ -n "$CI" ]]; then
@@ -73,8 +88,9 @@ Browser UAT Setup Check
 
 | Check              | Status | Details |
 |--------------------|--------|---------|
+| gsd-browser        | {Y|X}  | {path/version | not found} |
 | ANTHROPIC_API_KEY  | {Y|X}  | {set to sk-...xxx | not set} |
-| agent-browse       | {Y|X}  | {installed | not installed} |
+| agent-browse       | {Y|X}  | {installed | not installed; fallback only} |
 | Dev server         | {Y|!}  | {running at URL | not reachable} |
 | Environment        | {value}| {Interactive | Headless} |
 
@@ -89,33 +105,45 @@ Reason: {SPECIFIC_REASON}
 ## Tool Selection Logic
 
 ```
-if api_key_present AND plugin_installed:
+if gsd_browser_present:
+  tool = "gsd-browser"
+  reason = "GSD browser is installed and is the default for browser UAT"
+
+elif api_key_present AND plugin_installed:
   tool = "agent-browse"
-  reason = "API key found + plugin installed = full automation available"
+  reason = "GSD browser missing; API key + plugin available as fallback"
 
 elif api_key_present AND environment == "interactive":
   tool = "claude --chrome"
-  reason = "API key found but plugin missing. Falling back to interactive Chrome."
-  suggestion = "Install plugin for autonomous testing: /plugin marketplace add browserbase/agent-browse"
+  reason = "GSD browser missing and plugin missing. Falling back to interactive Chrome."
+  suggestion = "Install gsd-browser for default autonomous browser UAT"
 
 elif environment == "interactive":
   tool = "claude --chrome (manual)"
-  reason = "No API key. You'll need to test manually in Chrome."
-  suggestion = "Set ANTHROPIC_API_KEY for automated browser testing"
+  reason = "GSD browser missing and no API key. You'll need to test manually in Chrome."
+  suggestion = "Install gsd-browser for automated browser UAT"
 
 else:
   tool = "Skipped"
-  reason = "Headless mode without API key = no browser testing available"
-  suggestion = "Set ANTHROPIC_API_KEY=... to enable browser UAT in Pushpa/CI"
+  reason = "Headless mode without gsd-browser or API key = no browser testing available"
+  suggestion = "Install gsd-browser or set ANTHROPIC_API_KEY=... to enable fallback browser UAT in Pushpa/CI"
 ```
 
 ## Next Steps (conditional)
 
+**If tool = "gsd-browser":**
+```
+Ready for Browser UAT
+---------------------
+Your setup supports default browser testing with GSD Browser.
+No further action needed.
+```
+
 **If tool = "agent-browse":**
 ```
 Ready for Browser UAT
 ---------------------
-Your setup supports fully automated browser testing.
+Your setup supports fallback automated browser testing via agent-browse.
 No further action needed.
 ```
 
@@ -125,9 +153,8 @@ Fallback Mode: claude --chrome
 ------------------------------
 Browser UAT will work but requires you to be present.
 
-To upgrade to autonomous testing:
-  /plugin marketplace add browserbase/agent-browse
-  /plugin install agent-browse@browserbase
+To upgrade to default autonomous testing:
+  Install gsd-browser and ensure it is on PATH
 
 Then re-run: /rrr:check-browser-uat
 ```
@@ -138,8 +165,8 @@ Manual Mode: claude --chrome
 ----------------------------
 Browser UAT will prompt you to test manually.
 
-To enable AI-powered testing:
-  export ANTHROPIC_API_KEY=your-key-here
+To enable automated testing:
+  Install gsd-browser and ensure it is on PATH
 
 Then re-run: /rrr:check-browser-uat
 ```
@@ -151,10 +178,10 @@ Browser UAT will be SKIPPED
 Running in headless mode without ANTHROPIC_API_KEY.
 
 To enable browser UAT in Pushpa/CI:
-  1. export ANTHROPIC_API_KEY=your-key-here
-  2. Ensure agent-browse plugin is installed
+  1. Install gsd-browser and ensure it is on PATH
+  2. Or export ANTHROPIC_API_KEY and ensure agent-browse is installed as fallback
 
-Browser UAT is optional - Playwright tests still run.
+Browser UAT is optional - deterministic tests still run.
 See: rrr/references/browser-testing.md
 ```
 

package/commands/rrr/check-upstream.md

@@ -1,6 +1,6 @@
 # Check for RRR-impacting upstream changes
 
-Runs all upstream watchers (Claude Code, Get-Shit-Done, Taches CC Resources) and reports changes that may affect RRR workflows.
+Runs all upstream watchers (Claude Code, Get-Shit-Done, GSD Browser, Superpowers, Vercel Agent Browser, Taches CC Resources) and reports changes that may affect RRR workflows.
 
 ## Usage
 
@@ -27,6 +27,9 @@ MEDIUM risk changes should be reviewed for improvement opportunities.
 |----------|------|------------------|
 | Claude Code | toolchain | CLI changes affecting RRR workflows |
 | Get-Shit-Done | strict_baseline | Commands, agents, scripts that RRR extends |
+| GSD Browser | browser_uat_baseline | Browser automation CLI used by RRR UAT |
+| Superpowers | methodology_harvest | Skills/workflow patterns for disciplined agent execution |
+| Vercel Agent Browser | browser_capability_harvest | Browser automation capabilities worth adapting |
 | Taches CC Resources | capability_harvest | Skills, prompts, workflows for pattern harvest |
 
 ## Examples

package/commands/rrr/coordinate-merge.md

@@ -0,0 +1,70 @@
+---
+name: rrr:coordinate-merge
+description: Generate a GitHub-native coordinator merge readiness report
+argument-hint: "[--dry-run]"
+allowed-tools:
+  - Read
+  - Write
+  - Bash
+  - AskUserQuestion
+---
+
+<objective>
+Act as the coordinator merge reviewer for RRR team mode.
+
+This command is report-first. It should not merge automatically in pilot mode. It inspects team submissions, GitHub-native readiness signals when available, and produces the final integration report for the human coordinator.
+</objective>
+
+<process>
+
+1. **Load team-mode state**
+   - Read `.planning/team-mode.json`.
+   - Read `.planning/WORKSTREAMS.md`.
+   - Read `.planning/workstreams/*/SUBMISSION.json`.
+
+2. **Generate local report**
+   Run:
+
+   ```bash
+   node ~/.claude/rrr/scripts/rrr-team-mode.js report
+   ```
+
+   Development source repo fallback:
+
+   ```bash
+   node scripts/rrr-team-mode.js report
+   ```
+
+3. **GitHub-native checks**
+   If GitHub credentials/tools are available, inspect:
+   - Assigned phase issues
+   - Team PRs
+   - PR changed files
+   - Checks
+   - CODEOWNER reviews
+   - Mergeability against integration branch
+
+4. **Integration simulation**
+   If coordinator runner/Sprite is available:
+   - Checkout integration branch
+   - Merge team branches in proposed order
+   - Run build/test/UAT commands
+   - Record command output and artifacts
+
+5. **Report**
+   Update `.planning/integration/REPORT.md` with:
+   - ready/blocking teams
+   - incomplete phases
+   - files touched by each team
+   - cross-team conflict risks
+   - missing reviews/checks
+   - proposed merge order
+   - final recommendation
+
+</process>
+
+<constraints>
+- Do not auto-merge in pilot mode.
+- Do not treat local RRR state as stronger than GitHub PR/check/review state.
+- Surface unknowns explicitly; do not convert missing data into a pass.
+</constraints>

package/commands/rrr/dispatch-team.md

@@ -0,0 +1,57 @@
+---
+name: rrr:dispatch-team
+description: Prepare a GitHub team workstream for execution on a branch and Sprite
+argument-hint: "--team <team> [--dry-run]"
+allowed-tools:
+  - Read
+  - Bash
+  - AskUserQuestion
+---
+
+<objective>
+Dispatch an assigned team workstream to execution.
+
+Pilot behavior is report-first: verify planned branch, draft PR, and Sprite names, then give the coordinator exact commands or API actions to run. Live GitHub/Sprites mutation should happen only after explicit confirmation.
+</objective>
+
+<process>
+
+1. **Load assignment**
+   - Read `.planning/team-mode.json`.
+   - Read `.planning/workstreams/<team>/ASSIGNMENT.md`.
+   - Run:
+
+   ```bash
+   node ~/.claude/rrr/scripts/rrr-team-mode.js status
+   ```
+
+2. **Preflight GitHub-native contract**
+   Verify the team has:
+   - Assigned phases
+   - GitHub team name
+   - Branch name
+   - Sprite name
+   - Base branch and integration branch
+
+3. **Dispatch checklist**
+   Present the actions:
+   - Create or update branch `rrr/<milestone>/<team>`
+   - Create draft PR into `integration/<milestone>`
+   - Add labels: `rrr`, `team:<team>`, `milestone:<milestone>`
+   - Link assigned GitHub phase issues
+   - Request team review from the GitHub team
+   - Create or wake the configured Sprite
+   - Clone repo and checkout assigned branch
+   - Run `$rrr-next --team <team>` inside Sprite
+
+4. **Confirmation**
+   Ask before making any external mutations.
+   In pilot mode, if credentials are not configured, stop after the checklist.
+
+</process>
+
+<constraints>
+- Do not use shared PATs. Prefer GitHub App installation tokens for automation.
+- Do not treat a Sprite shell session as an isolation boundary; use one Sprite per team/workstream.
+- Do not auto-merge or auto-mark work ready.
+</constraints>

package/commands/rrr/execute-phase.md

@@ -1,7 +1,7 @@
 ---
 name: rrr:execute-phase
 description: Execute all plans in a phase with wave-based parallelization
-argument-hint: "<phase-number> [--auto] [--dry-run] [--step-cap=30m] [--pipeline-cap=4h] [--wall-cap=6h]"
+argument-hint: "<phase-number> [--team <team>] [--auto] [--dry-run] [--step-cap=30m] [--pipeline-cap=4h] [--wall-cap=6h]"
 allowed-tools:
   - Read
   - Write
@@ -22,6 +22,8 @@ Orchestrator stays lean: discover plans, analyze dependencies, group into waves,
 Context budget: ~15% orchestrator, 100% fresh per subagent.
 
 When `--auto` flag is passed, chains discuss-plan-execute automatically for the phase (and subsequent phases if applicable). Pipeline enforces triple budget caps and checks for `.planning/.stop-auto` stop-file before each step. Use `--dry-run` with `--auto` to preview the pipeline plan without executing.
+
+When `--team <team>` is passed, execute as a team workstream command: root `.planning/STATE.md` is read-only coordinator context, `.planning/workstreams/<team>/STATE.md` is the writable execution state, and root roadmap/state completion updates are deferred to coordinator integration.
 </objective>
 
 <execution_context>
@@ -30,6 +32,7 @@ When `--auto` flag is passed, chains discuss-plan-execute automatically for the
 @~/.claude/rrr/workflows/execute-phase.md
 @rrr/lib/phase-paths.md
 @rrr/references/semantic-search-preference.md
+@rrr/references/team-mode.md
 </execution_context>
 
 <context>
@@ -43,6 +46,16 @@ Phase: $ARGUMENTS
 
 <process>
 
+0a. **Parse team mode**
+   If `$ARGUMENTS` contains `--team <team>`:
+   - Set `TEAM_MODE=true` and `TEAM=<team>`.
+   - Read `.planning/team-mode.json`.
+   - Read `.planning/workstreams/<team>/STATE.md`.
+   - Confirm the requested phase is assigned to that team, or ask the coordinator before proceeding.
+   - Treat `.planning/STATE.md` as read-only coordinator context.
+   - Any step below that says to update or stage `.planning/STATE.md` or `.planning/ROADMAP.md` must instead update `.planning/workstreams/<team>/STATE.md` and leave root state/roadmap untouched.
+   - Final readiness is recorded by `/rrr:submit-phase <phase> --team <team>`.
+
 0. **Refresh scope cache (first)**
    ```bash
    node ~/.claude/rrr/scripts/refresh-scope-cache.js 2>/dev/null || echo "Cache refresh skipped"
@@ -278,14 +291,15 @@ Phase: $ARGUMENTS
    c. **Run browser UAT** (if UI_AFFECTING):
       ```bash
       # Run browser UAT via browser-uat.sh
-      # Handles: agent-browse (preferred) → claude --chrome (fallback) → skip (Pushpa/CI without key)
+      # Handles: gsd-browser (default) → agent-browse (fallback) → claude --chrome (interactive fallback) → skip
       FRONTEND_IMPACT=true PLAN_ID={phase} bash scripts/browser-uat.sh
       ```
 
       **Tool selection:**
-      - **ANTHROPIC_API_KEY set:** Uses agent-browse (works in Pushpa/CI)
-      - **No key + interactive:** Falls back to claude --chrome
-      - **No key + Pushpa/CI:** Skipped (no automated browser available)
+      - **gsd-browser available:** Uses gsd-browser (default, works in Pushpa/CI)
+      - **No gsd-browser + ANTHROPIC_API_KEY set:** Uses agent-browse fallback
+      - **No unattended tool + interactive:** Falls back to claude --chrome
+      - **No unattended tool + Pushpa/CI:** Skipped
 
    d. **Confirm artifact locations** (if tests ran):
       - `.planning/artifacts/playwright/test-results/` — screenshots, traces, videos
@@ -301,7 +315,7 @@ Phase: $ARGUMENTS
       **Commands:**
       - `npm test` — {pass/fail or "skipped"}
       - `npx playwright test` — {pass/fail or "skipped" or "n/a (backend_only)"}
-      - `browser_uat` — {pass/fail or "skipped" or "n/a (backend_only)"} (tool: agent-browse | claude-chrome)
+      - `browser_uat` — {pass/fail or "skipped" or "n/a (backend_only)"} (tool: gsd-browser | agent-browse | claude-chrome)
 
       **Result:** {PASS|FAIL} ({passed}/{total} steps)
 
@@ -320,7 +334,7 @@ Phase: $ARGUMENTS
 
    | Surface | unit_tests | playwright | browser_uat |
    |---------|------------|------------|-------------|
-   | ui_affecting | Yes | Yes | Yes (agent-browse or claude --chrome) |
+   | ui_affecting | Yes | Yes | Yes (gsd-browser default; fallbacks available) |
    | backend_only | Yes | No | No |
 
    **Visual proof failure does NOT block phase completion** — logged as warning only.