crewly 1.2.1 → 1.2.4

package/config/skills/agent/computer-use/lib/playwright.sh

@@ -0,0 +1,153 @@
+#!/usr/bin/env bash
+# =============================================================================
+# Playwright CDP Integration — Layer 4
+# Connect to Chrome via Chrome DevTools Protocol for web app automation.
+# Requires Chrome launched with --remote-debugging-port.
+# =============================================================================
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "${SCRIPT_DIR}/../../_common/lib.sh"
+
+DEFAULT_CDP_PORT=9222
+
+# -----------------------------------------------------------------------------
+# chrome_connect — Test CDP connection to Chrome
+# Params: port (default 9222)
+# -----------------------------------------------------------------------------
+chrome_connect() {
+  local port="${1:-$DEFAULT_CDP_PORT}"
+
+  # Check if Chrome is reachable via CDP
+  local response
+  response=$(curl -s --connect-timeout 3 "http://127.0.0.1:${port}/json/version" 2>&1) || {
+    echo "{\"success\":false,\"action\":\"chrome-connect\",\"port\":$port,\"error\":\"Cannot connect to Chrome CDP on port $port. Launch Chrome with: open -a 'Google Chrome' --args --remote-debugging-port=$port\"}"
+    return 1
+  }
+
+  local browser
+  browser=$(echo "$response" | jq -r '.Browser // "unknown"' 2>/dev/null || echo "unknown")
+  local ws_url
+  ws_url=$(echo "$response" | jq -r '.webSocketDebuggerUrl // "unknown"' 2>/dev/null || echo "unknown")
+
+  echo "{\"success\":true,\"action\":\"chrome-connect\",\"port\":$port,\"browser\":\"$browser\",\"wsUrl\":\"$ws_url\"}"
+}
+
+# -----------------------------------------------------------------------------
+# chrome_tabs — List open Chrome tabs via CDP
+# Params: port (default 9222)
+# -----------------------------------------------------------------------------
+chrome_tabs() {
+  local port="${1:-$DEFAULT_CDP_PORT}"
+
+  local response
+  response=$(curl -s --connect-timeout 3 "http://127.0.0.1:${port}/json/list" 2>&1) || {
+    echo "{\"success\":false,\"action\":\"chrome-tabs\",\"error\":\"Cannot connect to Chrome CDP on port $port\"}"
+    return 1
+  }
+
+  # Filter to page type only and extract useful fields
+  local tabs
+  tabs=$(echo "$response" | jq '[.[] | select(.type == "page") | {id: .id, title: .title, url: .url}]' 2>/dev/null || echo "[]")
+
+  local count
+  count=$(echo "$tabs" | jq 'length' 2>/dev/null || echo "0")
+
+  echo "{\"success\":true,\"action\":\"chrome-tabs\",\"port\":$port,\"count\":$count,\"tabs\":$tabs}"
+}
+
+# -----------------------------------------------------------------------------
+# chrome_eval — Run JavaScript in a Chrome tab via CDP
+# Params: port, tabIndex, code
+# Uses the REST endpoint for simple evaluations.
+# -----------------------------------------------------------------------------
+chrome_eval() {
+  local port="${1:-$DEFAULT_CDP_PORT}"
+  local tab_index="${2:-0}"
+  local code="$3"
+
+  # Get tab list
+  local tabs_response
+  tabs_response=$(curl -s --connect-timeout 3 "http://127.0.0.1:${port}/json/list" 2>&1) || {
+    echo "{\"success\":false,\"action\":\"chrome-eval\",\"error\":\"Cannot connect to Chrome CDP\"}"
+    return 1
+  }
+
+  # Get the specific tab's webSocketDebuggerUrl
+  local ws_url
+  ws_url=$(echo "$tabs_response" | jq -r ".[$tab_index].webSocketDebuggerUrl // empty" 2>/dev/null)
+
+  if [ -z "$ws_url" ]; then
+    echo "{\"success\":false,\"action\":\"chrome-eval\",\"error\":\"Tab index $tab_index not found\"}"
+    return 1
+  fi
+
+  local tab_id
+  tab_id=$(echo "$tabs_response" | jq -r ".[$tab_index].id // empty" 2>/dev/null)
+
+  # Use CDP REST API to evaluate: activate tab first, then use the evaluate endpoint
+  # Note: For simple evaluations, we use osascript to run JS in Chrome via AppleScript
+  local tab_url
+  tab_url=$(echo "$tabs_response" | jq -r ".[$tab_index].url // empty" 2>/dev/null)
+  local tab_title
+  tab_title=$(echo "$tabs_response" | jq -r ".[$tab_index].title // empty" 2>/dev/null)
+
+  # Use the CDP WebSocket via a simple node script if available
+  if command -v node &>/dev/null; then
+    local result
+    result=$(node -e "
+      const http = require('http');
+      const data = JSON.stringify({
+        id: 1,
+        method: 'Runtime.evaluate',
+        params: { expression: $(echo "$code" | jq -Rs .), returnByValue: true }
+      });
+
+      // Use HTTP endpoint to send CDP command
+      const ws = require('ws');
+      const client = new ws('$ws_url');
+      let timeout = setTimeout(() => {
+        console.log(JSON.stringify({success: false, error: 'Timeout'}));
+        process.exit(1);
+      }, 10000);
+
+      client.on('open', () => {
+        client.send(data);
+      });
+      client.on('message', (msg) => {
+        clearTimeout(timeout);
+        const resp = JSON.parse(msg);
+        if (resp.result && resp.result.result) {
+          console.log(JSON.stringify({
+            success: true,
+            action: 'chrome-eval',
+            tabId: '$tab_id',
+            value: resp.result.result.value,
+            type: resp.result.result.type
+          }));
+        } else if (resp.result && resp.result.exceptionDetails) {
+          console.log(JSON.stringify({
+            success: false,
+            action: 'chrome-eval',
+            error: resp.result.exceptionDetails.text || 'Evaluation error'
+          }));
+        }
+        client.close();
+      });
+      client.on('error', (err) => {
+        clearTimeout(timeout);
+        console.log(JSON.stringify({success: false, error: err.message}));
+        process.exit(1);
+      });
+    " 2>&1) || {
+      # Fallback if ws module not available
+      echo "{\"success\":false,\"action\":\"chrome-eval\",\"error\":\"Node.js 'ws' module required. Install: npm install -g ws\"}"
+      return 1
+    }
+    echo "$result"
+  else
+    echo "{\"success\":false,\"action\":\"chrome-eval\",\"error\":\"Node.js required for chrome-eval. Install Node.js first.\"}"
+    return 1
+  fi
+}

package/config/skills/agent/computer-use/lib/screenshot.sh

@@ -0,0 +1,61 @@
+#!/usr/bin/env bash
+# =============================================================================
+# Screenshot Utility — captures screen or specific app windows
+# =============================================================================
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "${SCRIPT_DIR}/../../_common/lib.sh"
+
+# -----------------------------------------------------------------------------
+# take_screenshot — Capture the screen or a specific app window
+# Params: output (file path), appName (optional, for window capture)
+# -----------------------------------------------------------------------------
+take_screenshot() {
+  local output="${1:-/tmp/screenshot.png}"
+  local app_name="${2:-}"
+
+  mkdir -p "$(dirname "$output")"
+
+  if [ -n "$app_name" ]; then
+    # Get the window ID for the app, then capture that window
+    local window_id
+    window_id=$(osascript -l JavaScript -e "
+      var app = Application('System Events');
+      var proc = app.processes.byName('$app_name');
+      var wins = proc.windows();
+      if (wins.length > 0) {
+        // Get the window number via attributes
+        try {
+          var attrs = wins[0].attributes();
+          for (var i = 0; i < attrs.length; i++) {
+            if (attrs[i].name() === 'AXWindow') {
+              // Fall through to full screen capture with app focus
+            }
+          }
+        } catch(e) {}
+      }
+      '';
+    " 2>/dev/null || echo "")
+
+    # Focus the app first, then capture the frontmost window
+    osascript -e "tell application \"$app_name\" to activate" 2>/dev/null || true
+    sleep 0.5
+    screencapture -x -o -l "$(osascript -e 'tell application "System Events" to get id of first window of first process whose frontmost is true' 2>/dev/null || echo "")" "$output" 2>/dev/null || {
+      # Fallback: capture full screen
+      screencapture -x "$output" 2>/dev/null
+    }
+  else
+    # Full screen capture
+    screencapture -x "$output" 2>/dev/null
+  fi
+
+  if [ -f "$output" ]; then
+    local size
+    size=$(stat -f%z "$output" 2>/dev/null || stat -c%s "$output" 2>/dev/null || echo "0")
+    echo "{\"success\":true,\"action\":\"screenshot\",\"file\":\"$output\",\"size\":$size,\"appName\":\"${app_name:-full-screen}\"}"
+  else
+    error_exit "Screenshot failed — file not created"
+  fi
+}

package/config/skills/agent/computer-use/skill.json

@@ -0,0 +1,29 @@
+{
+  "id": "computer-use",
+  "name": "Agent Browser — Universal Desktop Control",
+  "description": "Universal desktop application control for macOS. Integrates three control methods: AppleScript/JXA for native app automation, Accessibility API for universal UI reading/interaction, and Playwright CDP for Chrome web apps. Includes app discovery, safety logging, and screenshot auditing.",
+  "category": "automation",
+  "skillType": "claude-skill",
+  "promptFile": "instructions.md",
+  "execution": {
+    "type": "script",
+    "script": {
+      "file": "execute.sh",
+      "interpreter": "bash",
+      "timeoutMs": 60000
+    }
+  },
+  "triggers": [
+    "computer use", "desktop automation", "browser", "screenshot",
+    "click", "type text", "applescript", "accessibility",
+    "ui tree", "chrome", "playwright", "control app"
+  ],
+  "tags": [
+    "automation", "desktop", "browser", "screenshot", "applescript",
+    "accessibility", "playwright", "macos", "ui-reading", "cdp"
+  ],
+  "assignableRoles": ["developer", "generalist", "designer", "qa"],
+  "version": "1.0.0",
+  "createdAt": "2026-03-06T00:00:00.000Z",
+  "updatedAt": "2026-03-06T00:00:00.000Z"
+}

package/config/skills/agent/core/accept-task/execute.sh

@@ -11,11 +11,17 @@ SESSION_NAME=$(echo "$INPUT" | jq -r '.sessionName // empty')
 require_param "sessionName" "$SESSION_NAME"
 
 TEAM_MEMBER_ID=$(echo "$INPUT" | jq -r '.teamMemberId // empty')
+PROJECT_PATH=$(echo "$INPUT" | jq -r '.projectPath // empty')
+TASK_GROUP=$(echo "$INPUT" | jq -r '.taskGroup // empty')
 
 BODY=$(jq -n \
   --arg sessionName "$SESSION_NAME" \
   --arg teamMemberId "$TEAM_MEMBER_ID" \
+  --arg projectPath "$PROJECT_PATH" \
+  --arg taskGroup "$TASK_GROUP" \
   '{sessionName: $sessionName} +
-   (if $teamMemberId != "" then {teamMemberId: $teamMemberId} else {} end)')
+   (if $teamMemberId != "" then {teamMemberId: $teamMemberId} else {} end) +
+   (if $projectPath != "" then {projectPath: $projectPath} else {} end) +
+   (if $taskGroup != "" then {taskGroup: $taskGroup} else {} end)')
 
 api_call POST "/task-management/take-next" "$BODY"

package/config/skills/agent/core/complete-task/execute.sh

@@ -16,14 +16,51 @@ require_param "absoluteTaskPath" "$ABSOLUTE_TASK_PATH"
 require_param "sessionName" "$SESSION_NAME"
 require_param "summary" "$SUMMARY"
 
+# Optional structured VerificationRequest fields (for hierarchical workflows)
+TASK_ID=$(echo "$INPUT" | jq -r '.taskId // empty')
+ARTIFACTS=$(echo "$INPUT" | jq -c '.artifacts // empty')
+TEST_RESULTS=$(echo "$INPUT" | jq -r '.testResults // empty')
+USE_STRUCTURED=$(echo "$INPUT" | jq -r '.structured // "false"')
+
+# If structured mode is enabled and taskId is provided, send a [VERIFICATION REQUEST]
+# to the orchestrator/team-leader before completing the task file.
+if [ "$USE_STRUCTURED" = "true" ] && [ -n "$TASK_ID" ]; then
+  VER_MESSAGE="---\n[VERIFICATION REQUEST]\nTask ID: ${TASK_ID}\nRequested by: ${SESSION_NAME}\n---\n\n## Summary\n${SUMMARY}"
+
+  # Add artifacts if provided
+  if [ -n "$ARTIFACTS" ] && [ "$ARTIFACTS" != "" ]; then
+    ARTIFACT_LINES=$(echo "$ARTIFACTS" | jq -r '.[]? | "- **\(.name)** (\(.type)): \(.content)"' 2>/dev/null || true)
+    if [ -n "$ARTIFACT_LINES" ]; then
+      VER_MESSAGE="${VER_MESSAGE}\n\n## Artifacts\n${ARTIFACT_LINES}"
+    fi
+  fi
+
+  # Add test results if provided
+  if [ -n "$TEST_RESULTS" ]; then
+    VER_MESSAGE="${VER_MESSAGE}\n\n## Test Results\n${TEST_RESULTS}"
+  fi
+
+  # Send verification request to orchestrator via chat API
+  VER_BODY=$(jq -n --arg content "$VER_MESSAGE" --arg senderName "$SESSION_NAME" \
+    '{content: $content, senderName: $senderName, senderType: "agent"}')
+  api_call POST "/chat/agent-response" "$VER_BODY" || true
+fi
+
 BODY=$(jq -n \
   --arg absoluteTaskPath "$ABSOLUTE_TASK_PATH" \
+  --arg taskPath "$ABSOLUTE_TASK_PATH" \
   --arg sessionName "$SESSION_NAME" \
   --arg summary "$SUMMARY" \
   --arg skipGates "$SKIP_GATES" \
   --argjson output "${OUTPUT_JSON:-null}" \
-  '{absoluteTaskPath: $absoluteTaskPath, sessionName: $sessionName, summary: $summary} +
+  '{absoluteTaskPath: $absoluteTaskPath, taskPath: $taskPath, sessionName: $sessionName, summary: $summary} +
    (if $skipGates == "true" then {skipGates: true} else {} end) +
    (if $output != null then {output: $output} else {} end)')
 
 api_call POST "/task-management/complete" "$BODY"
+
+# Auto-persist the task summary as project knowledge (#127).
+if [ -n "$SUMMARY" ]; then
+  PROJECT_PATH=$(echo "$INPUT" | jq -r '.projectPath // empty')
+  auto_remember "$SESSION_NAME" "Task completed by ${SESSION_NAME}: ${SUMMARY}" "pattern" "project" "$PROJECT_PATH"
+fi

package/config/skills/agent/core/report-status/execute.sh

@@ -15,9 +15,52 @@ require_param "sessionName" "$SESSION_NAME"
 require_param "status" "$STATUS"
 require_param "summary" "$SUMMARY"
 
+# Optional structured StatusReport fields (for hierarchical workflows)
+TASK_ID=$(echo "$INPUT" | jq -r '.taskId // empty')
+PROGRESS=$(echo "$INPUT" | jq -r '.progress // empty')
+ARTIFACTS=$(echo "$INPUT" | jq -c '.artifacts // empty')
+BLOCKERS=$(echo "$INPUT" | jq -c '.blockers // empty')
+USE_STRUCTURED=$(echo "$INPUT" | jq -r '.structured // "false"')
+
+# Map simple status strings to InProgressTaskStatus values
+map_status_to_state() {
+  case "$1" in
+    done|completed) echo "completed" ;;
+    blocked) echo "blocked" ;;
+    failed) echo "failed" ;;
+    in_progress|working) echo "working" ;;
+    *) echo "$1" ;;
+  esac
+}
+
 # Build the message the orchestrator will receive
-STATUS_UPPER=$(echo "$STATUS" | tr '[:lower:]' '[:upper:]')
-MESSAGE="[${STATUS_UPPER}] Agent ${SESSION_NAME}: ${SUMMARY}"
+if [ "$USE_STRUCTURED" = "true" ] && [ -n "$TASK_ID" ]; then
+  # Structured StatusReport format for hierarchical teams
+  STATE=$(map_status_to_state "$STATUS")
+  MESSAGE="---\n[STATUS REPORT]\nTask ID: ${TASK_ID}\nState: ${STATE}"
+  [ -n "$PROGRESS" ] && MESSAGE="${MESSAGE}\nProgress: ${PROGRESS}%"
+  MESSAGE="${MESSAGE}\nReported by: ${SESSION_NAME}\n---\n\n## Status\n${SUMMARY}"
+
+  # Add artifacts if provided
+  if [ -n "$ARTIFACTS" ] && [ "$ARTIFACTS" != "" ]; then
+    ARTIFACT_LINES=$(echo "$ARTIFACTS" | jq -r '.[]? | "- **\(.name)** (\(.type)): \(.content)"' 2>/dev/null || true)
+    if [ -n "$ARTIFACT_LINES" ]; then
+      MESSAGE="${MESSAGE}\n\n## Artifacts\n${ARTIFACT_LINES}"
+    fi
+  fi
+
+  # Add blockers if provided
+  if [ -n "$BLOCKERS" ] && [ "$BLOCKERS" != "" ]; then
+    BLOCKER_LINES=$(echo "$BLOCKERS" | jq -r '.[]? // empty' 2>/dev/null | while read -r b; do echo "- ${b}"; done)
+    if [ -n "$BLOCKER_LINES" ]; then
+      MESSAGE="${MESSAGE}\n\n## Blockers\n${BLOCKER_LINES}"
+    fi
+  fi
+else
+  # Legacy format (backwards compatible)
+  STATUS_UPPER=$(echo "$STATUS" | tr '[:lower:]' '[:upper:]')
+  MESSAGE="[${STATUS_UPPER}] Agent ${SESSION_NAME}: ${SUMMARY}"
+fi
 
 # Send the status message to the orchestrator session via the chat API
 BODY=$(jq -n --arg content "$MESSAGE" --arg senderName "$SESSION_NAME" \
@@ -39,3 +82,9 @@ if [ "$STATUS" = "done" ]; then
   SESSION_BODY=$(jq -n --arg sessionName "$SESSION_NAME" '{sessionName: $sessionName}')
   api_call POST "/task-management/complete-by-session" "$SESSION_BODY" || true
 fi
+
+# Auto-persist key findings as project knowledge when task is done (#127).
+if [ "$STATUS" = "done" ] && [ -n "$SUMMARY" ]; then
+  PROJECT_PATH=$(echo "$INPUT" | jq -r '.projectPath // empty')
+  auto_remember "$SESSION_NAME" "Task completed by ${SESSION_NAME}: ${SUMMARY}" "pattern" "project" "$PROJECT_PATH"
+fi