agentlife 2.6.10 → 2.6.11

package/dist/index.js

@@ -1129,28 +1129,9 @@ context: <JSON — lookup keys: DB tables, date ranges, phase, config>
 
 ### Two-Phase Push — Loading Then Final
 
-Phase 1 — your FIRST tool call on every turn. Before any search, fetch, exec, or db call:
+Phase 1 — your FIRST tool call on every turn. Before any search, fetch, exec, or db call, push a widget with \`state=loading\` in the header. That attribute is the only rule: it triggers the animated border on the dashboard so the user sees "work in progress." Everything else — size, layout, what the face shows, even whether it's a brand-new surface or a reused one — is your call. Match the feel of the domain: a data fetch can show a progress bar, a long exec can list steps, a quick lookup can be one line.
 
-\`\`\`
-surface <id> size=s state=loading
-  card
-    column
-      text "<short description of what you are doing>" body
-\`\`\`
-
-Phase 2 — your LAST tool call. Same surfaceId, without state=loading:
-
-\`\`\`
-surface <id> size=m
-  card
-    column
-      <face: metrics, badges, gauges — 2-second glance only>
-goal: <the user's objective — NOT the action you performed>
-detail:
-  <full markdown: paragraphs, tables, lists, analysis>
-followup: +<duration> "<next step — specific action based on current data>"
-context: {"key": "value"}
-\`\`\`
+Phase 2 — your LAST tool call. Same surfaceId, without \`state=loading\`, with the real result (face, goal, detail, followup, context). See TOOLS.md for DSL structure.
 
 A loading push without a final push = perpetual spinner.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentlife",
-  "version": "2.6.10",
+  "version": "2.6.11",
   "type": "module",
   "scripts": {
     "build": "bun build index.ts --outfile dist/index.js --target node --external openclaw/plugin-sdk",