ur-agent 1.40.1 → 1.41.1

package/docs/CONFIGURATION.md

@@ -153,9 +153,11 @@ never runs automatically unless enabled.
 
 ## Prompt Planning
 
-UR-Nexus can plan an `ur exec` prompt into small executable tasks, show a task
-board, run independent tasks through parallel logical workers, and verify task
-claims after execution. The defaults are:
+UR-Nexus can plan an `ur exec` prompt into ordered executable tasks, show a task
+board, run independent tasks through adaptive parallel logical workers, and
+verify task claims after execution. Short prompts stay as one compact task when
+splitting is not useful. Longer prompts are decomposed by explicit ordering,
+bullets, dependencies, and file targets. The defaults are:
 
 ```json
 {
@@ -168,9 +170,15 @@ claims after execution. The defaults are:
 ```
 
 `taskPlanning` enables prompt decomposition. `parallelAgents` allows independent
-tasks to run concurrently up to `maxAgents`. `showTaskBoard` renders visible
-progress during real execution and keeps the final board in the execution
-report. `strictVerification` rejects unsupported claims about changed files,
+tasks to run concurrently up to `maxAgents`; the scheduler still uses only the
+number of agents that is useful for the current dependency graph and file locks.
+Simple prompts use one agent, medium independent prompts use two or three when
+useful, and large independent task graphs can use the configured maximum.
+`showTaskBoard` renders visible progress during real execution and keeps the
+final ordered board in the execution report. The board shows ordered tasks,
+current status, the running task, finished tasks, tasks waiting for approval or
+context, active/max agents, queued tasks, and finished/failed/waiting/skipped
+counts. `strictVerification` rejects unsupported claims about changed files,
 commands, or generated output. With `--no-strict-verification`, unsupported
 claims become warnings and the task may finish when no hard execution error was
 observed.
@@ -189,12 +197,31 @@ ur exec "run quietly but keep the final board" --quiet
 ur exec "warn instead of failing unsupported claims" --no-strict-verification
 ```
 
+Risky actions use an approval-first workflow. UR-Nexus asks for approval before
+destructive commands, outside-workspace writes/deletes, network actions,
+credential-sensitive access, security testing commands, exploit-like commands,
+or commands that affect external systems. The request explains the action, why
+approval is required, and the command or file path when available. The action is
+not executed until approval evidence exists, and the decision is recorded in
+the final report.
+
+Outside-workspace reads are allowed when explicitly requested or clearly needed
+for the task, and the outside path is recorded as evidence. Modifying, removing,
+or deleting anything outside the workspace requires explicit approval first.
+For cybersecurity or security-research tasks, UR-Nexus supports authorized
+workflows by asking for target scope and authorization confirmation when needed.
+Vague requests are converted into scoped research tasks that wait for approval;
+local, lab, and test targets are preferred unless the user confirms authorized
+external scope.
+
 The final `ur exec` report is generated from task execution evidence only:
-finished, failed, and blocked task records, actual changed files from workspace
-snapshots, unreported changed files, verified commands surfaced by the executor,
-unverified command claims, verification failures, and warnings. Command tracking
-cannot prove detached or provider-internal activity unless the task runner
-surfaces those commands as observed evidence.
+finished, waiting approval/context, failed, and skipped task records, actual
+changed files from workspace snapshots, unreported changed files,
+outside-workspace files accessed or modified, verified commands surfaced by the
+executor, unverified command claims, approval decisions, verification failures,
+warnings, and remaining limitations. Command tracking cannot prove detached or
+provider-internal activity unless the task runner surfaces those commands as
+observed evidence.
 
 ## Project Safety Policy
 
@@ -338,6 +365,15 @@ Sandbox behavior has three modes, controlled by `sandbox.enabled` and
   UR fails closed: it refuses to start rather than run without a working
   sandbox.
 
+To turn the sandbox on or off, use the `/config` tool or run:
+
+```sh
+ur config set sandbox.enabled true
+ur config set sandbox.enabled false
+```
+
+You can also inspect the current state with `ur sandbox status`.
+
 OS confinement depends on platform support: `sandbox-exec` (Seatbelt) on
 macOS, or `bwrap` (bubblewrap) on Linux/WSL2. `ur sandbox check` reports
 missing dependencies for the current platform.

package/documentation/index.html

@@ -44,7 +44,7 @@
     <main id="content" class="content">
       <header class="topbar">
         <div>
-          <p class="eyebrow">Version 1.40.1</p>
+          <p class="eyebrow">Version 1.41.1</p>
           <h1>UR-Nexus Documentation</h1>
           <p class="lead">A practical, tutorial-style reference for installing, configuring, automating, extending, and operating UR-Nexus.</p>
         </div>

package/extensions/vscode-ur-inline-diffs/package.json

@@ -2,7 +2,7 @@
   "name": "ur-inline-diffs",
   "displayName": "UR Inline Diffs",
   "description": "Review, apply, and reject UR inline diff bundles from .ur/ide/diffs inside VS Code.",
-  "version": "1.40.1",
+  "version": "1.41.1",
   "publisher": "ur-nexus",
   "engines": {
     "vscode": "^1.92.0"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ur-agent",
-  "version": "1.40.1",
+  "version": "1.41.1",
   "description": "UR-Nexus — autonomous engineering workflow engine (plan, execute, test, verify, document, benchmark, reproduce)",
   "type": "module",
   "packageManager": "bun@1.3.14",