ur-agent 1.65.12 → 1.65.14

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -1,5 +1,47 @@
1
1
  # Changelog
2
2
 
3
+ ## 1.65.14
4
+
5
+ - Restored the proactive task-first behavior that was present in v1.65.0 and
6
+ removed a prompt/gate contradiction introduced later. For any non-trivial
7
+ state change, including a feature-rich single-file build, the model is now
8
+ told at system, task-tool, approved-plan, Ollama/Kimi, file, notebook,
9
+ shell, and worker boundaries to complete the available task setup, inspect
10
+ its result, mark the selected task `in_progress`, inspect that result, and
11
+ only then mutate state.
12
+ - Made the ordering unambiguous for weaker models: task setup cannot be batched
13
+ with the mutation it enables, approved-plan handoffs require task calls as
14
+ the next state-changing calls, terminal task lists must be reopened or
15
+ extended before new work, and task guidance appears before file-tool
16
+ selection guidance.
17
+ - Closed planner availability and control-flow gaps across tool modes. Legacy
18
+ `TodoWrite` now follows the same feature-rich/terminal-list lifecycle and is
19
+ preferred when Task V2 is incomplete; bare/simple, REPL-simple, coordinator,
20
+ custom-agent, and override-prompt paths all receive a usable planner and the
21
+ task-state contract. Coordinator workers must be task-bound before launch;
22
+ an explicitly planner-less custom tool pool now fails closed with honest
23
+ configuration recovery instead of naming a missing tool.
24
+ - Prevented task-gate deadlocks and false mutations in supporting tools.
25
+ Team creation/deletion, shutdown/plan control messages, and emergency task
26
+ stop are narrow control transitions; skill loading and desktop screenshots
27
+ are read-only while downstream skill work, desktop input, and arbitrary
28
+ state-changing tools remain gated.
29
+
30
+ ## 1.65.13
31
+
32
+ - Recovered otherwise valid `AskUserQuestion` calls whose UI header exceeds
33
+ the 12-character chip width. Native structured calls, bare/wrapped JSON, and
34
+ explicit-choice recovery now compact only that bounded presentation field;
35
+ questions, choices, labels, descriptions, previews, metadata, and selection
36
+ behavior remain unchanged, while unsafe or grossly oversized headers still
37
+ fail validation.
38
+ - Allowed strictly parsed syntax verification after a task-free one-shot
39
+ Write. `node --check <file>` and the bounded transcript-produced HTML script
40
+ checker bypass only the task-list requirement; arbitrary Node evaluation,
41
+ extra commands, redirects, backgrounding, sandbox overrides, and
42
+ permission-time rewrites fail closed, and normal Bash permission/sandbox
43
+ checks still apply.
44
+
3
45
  ## 1.65.12
4
46
 
5
47
  - Forced explicit weak-model decision menus through the real