ur-agent 1.81.0 → 1.81.2

package/docs/REDTEAM.md

@@ -61,4 +61,10 @@ Install/enable it from the official marketplace if it is not already enabled, ac
 /reverse-skills:start <research task>
 ```
 
+The command preflights UR's visible task lifecycle for multi-outcome or
+workspace-changing work. It creates an actionable board before Bash, edits,
+writes, or delegation, so `tasks.requireBeforeChanges` remains enabled without
+surfacing a failed-first `TaskListRequired` tool call. Direct specialist-skill
+invocations follow the same contract.
+
 Review [the plugin README](../plugins/core/reverse-skills/README.md) for its capability map and provenance.

package/docs/VALIDATION.md

@@ -19,7 +19,7 @@ You need:
 
 ```sh
 ur --version
-# expected for this release: "1.81.0 (UR-Nexus)"
+# expected for this release: "1.81.2 (UR-Nexus)"
 ```
 
 ### 0.0 Redteam mode and Reverse Skills (1.81.0)

package/docs/providers.md

@@ -146,6 +146,23 @@ OpenAI-compatible, OpenRouter, local, or subscription-CLI providers.
 
 UR-Nexus shows providers first, then only models available for the selected provider. This prevents incompatible model/provider pairs and keeps API-key, local/server, subscription, and external app bridge model lists separate. The generic `subscription` entry has no models unless a real independent subscription runtime is configured; UR does not list fake subscription models.
 
+The interactive selection is also the authority for the current session: the
+next main, compact, or web-search request uses the provider/model pair captured
+in live app state while the same validated pair is persisted once to local
+settings. A stale settings snapshot cannot combine a newly selected model with
+the previous provider.
+
+### Reasoning effort
+
+Use `/effort low|medium|high|max|auto` inside UR. `max` is provider-neutral and
+means the selected model's highest reasoning capability. For OpenRouter, UR
+preserves the live `/models` reasoning metadata and sends the unified
+`reasoning.effort` request, translating `max` to the model's advertised wire
+value such as `xhigh` or `high`. The command confirmation, status indicator,
+and active-work spinner all show the same applied UR effort level. Providers
+that cannot apply a requested level report the actual downgraded level instead
+of confirming a value they did not use.
+
 ## Runtime provider routing
 
 When you select a UR-native provider and model, every agent request is routed

package/documentation/index.html

@@ -45,7 +45,7 @@
     <main id="content" class="content">
       <header class="topbar">
         <div>
-          <p class="eyebrow">Version 1.81.0</p>
+          <p class="eyebrow">Version 1.81.2</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/jetbrains-ur/build.gradle.kts

@@ -7,7 +7,7 @@ plugins {
 }
 
 group = "dev.urnexus"
-version = "1.81.0"
+version = "1.81.2"
 
 repositories {
     mavenCentral()

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": "1.81.0",
+  "version": "1.81.2",
   "publisher": "ur-nexus",
   "engines": {
     "vscode": "^1.92.0"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ur-agent",
-  "version": "1.81.0",
+  "version": "1.81.2",
   "description": "UR-Nexus — autonomous engineering workflow engine (plan, execute, test, verify, document, benchmark, reproduce)",
   "type": "module",
   "packageManager": "bun@1.3.14",

package/plugins/core/reverse-skills/.ur-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "reverse-skills",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "UR-native reverse engineering and red-team research workflows, available exclusively in UR redteam mode.",
   "author": {
     "name": "UR maintainers",

package/plugins/core/reverse-skills/README.md

@@ -12,6 +12,12 @@ The integration is UR-only:
 
 Start with `/reverse-skills:start <task>` (or invoke the `reverse-skills:reverse-skill-router` skill directly). The router selects one or more specialist skills covering binary reverse engineering, exploitation, application testing, malware/EDR research, platform and radio security, forensics, LLM security, and evidence/reporting.
 
+For multi-outcome or workspace-changing research, the command and every
+specialist skill expose UR's task tools and require an actionable task board
+before Bash, edits, writes, or delegation. This keeps the normal
+`tasks.requireBeforeChanges` protection enabled without a failed-first
+`TaskListRequired` retry.
+
 ## Provenance
 
 The adaptation is based on upstream commit `289c24b1617411a16b1e8d3032cce0f2fe52911d` (MIT). UR excludes the upstream GPL CTF orchestrator, external AGPL services, client-specific bootstrap scripts, field journals, and machine-specific configuration. See `NOTICE` and `UPSTREAM-LICENSE`.

package/plugins/core/reverse-skills/UR-INTEGRATION.md

@@ -3,13 +3,14 @@
 These rules apply to every skill in this plugin.
 
 1. Confirm UR redteam mode is active. Plugin gating normally guarantees this; never suggest bypassing the gate.
-2. Separate offline research from target operations. Reading, decompiling, documenting, and authoring local research artifacts may proceed under normal file/tool permissions.
-3. Before network access, sample execution, exploitation, credential testing, radio transmission, device mutation, or any other target operation, inspect `/scope show`. Require a scope approved in the current UR session. Mode activation and warning acceptance are not target authorization.
-4. Use only the hosts, ports, tools, and intensity recorded in scope. Stop when a discovered pivot or target is outside it; ask the user to update and reapprove scope.
-5. Continue to honor UR permission prompts, sandboxing, secrets handling, and action-specific confirmations. Never weaken these controls or silently elevate privileges.
-6. Prefer reproducible case directories, hashes, timestamps, commands, raw evidence, and findings that distinguish observation from inference.
-7. Do not auto-install tools or alter another AI client's configuration. If a dependency is absent, report it and propose an UR-native installation or MCP setup for explicit approval.
-8. The selected model/provider is independent. If it declines a request, state that accurately; do not imply UR redteam mode can override provider policy.
+2. Keep UR's task lifecycle ahead of mutations. Read-only inspection may happen first. If the request has multiple outcomes, sequencing, verification/reporting deliverables, or any workspace-changing action, call `TaskList` and create 2–8 concrete outcome tasks with `TaskCreate` when no actionable board exists. Mark the active task `in_progress` before Bash/Edit/Write or delegation, and update every task through completion, failure, or an explicit skip. Never wait for a `TaskListRequired` rejection and never disable the task gate as a workflow shortcut.
+3. Separate offline research from target operations. Reading, decompiling, documenting, and authoring local research artifacts may proceed under normal file/tool permissions.
+4. Before network access, sample execution, exploitation, credential testing, radio transmission, device mutation, or any other target operation, inspect `/scope show`. Require a scope approved in the current UR session. Mode activation and warning acceptance are not target authorization.
+5. Use only the hosts, ports, tools, and intensity recorded in scope. Stop when a discovered pivot or target is outside it; ask the user to update and reapprove scope.
+6. Continue to honor UR permission prompts, sandboxing, secrets handling, action-specific confirmations, and task tracking. Never weaken these controls or silently elevate privileges.
+7. Prefer reproducible case directories, hashes, timestamps, commands, raw evidence, and findings that distinguish observation from inference.
+8. Do not auto-install tools or alter another AI client's configuration. If a dependency is absent, report it and propose an UR-native installation or MCP setup for explicit approval.
+9. The selected model/provider is independent. If it declines a request, state that accurately; do not imply UR redteam mode can override provider policy.
 
 Recommended case layout:
 

package/plugins/core/reverse-skills/commands/start.md

@@ -3,8 +3,18 @@ description: "Route an authorized security-research task into the UR-exclusive R
 argument-hint: "<research task>"
 allowed-tools:
   - "Skill"
+  - "TaskCreate"
+  - "TaskList"
+  - "TaskUpdate"
 ---
 
+Before invoking another skill or using Bash, Edit, or Write, inspect the request
+for multiple outcomes, sequencing, verification/reporting deliverables, or any
+workspace-changing action. When any of those apply, call `TaskList`, create
+2–8 concrete outcome tasks with `TaskCreate` if no actionable board exists,
+and mark the first unblocked task in progress with `TaskUpdate`. Do not wait for
+`TaskListRequired`, and do not disable UR's task gate to bypass this preflight.
+
 Invoke the `reverse-skills:reverse-skill-router` skill with the user's arguments.
 Follow its UR redteam execution contract, select the smallest applicable set of
 specialist skills, and begin with local/passive triage. Do not interpret redteam

package/plugins/core/reverse-skills/skills/application-redteam/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: application-redteam
 description: Conduct authorized application and infrastructure research across APIs, source code, web, cloud/Kubernetes, databases, identity federation, email, supply chain, browser automation, and thick clients.
-allowed-tools: Read Grep Glob Bash Edit Write WebFetch WebSearch
+allowed-tools: Read Grep Glob Bash Edit Write WebFetch WebSearch TaskCreate TaskList TaskUpdate
 ---
 
 # Application red-team research

package/plugins/core/reverse-skills/skills/binary-reverse/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: binary-reverse
 description: Reverse compiled, managed, packed, obfuscated, browser, bytecode, VM, or protocol artifacts using static and dynamic analysis. Covers IDA, Ghidra, radare2, .NET, Go/Rust, macOS, JavaScript, browser extensions, binary diffing, and protocol reconstruction.
-allowed-tools: Read Grep Glob Bash Edit Write
+allowed-tools: Read Grep Glob Bash Edit Write TaskCreate TaskList TaskUpdate
 ---
 
 # Binary reverse engineering

package/plugins/core/reverse-skills/skills/exploit-development/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: exploit-development
 description: Turn a known vulnerability into a reproducible PoC or reliable authorized exploit, including stack/heap/kernel pwn, patch-diff analysis, exploit chains, fuzzing, and remote-environment stabilization.
-allowed-tools: Read Grep Glob Bash Edit Write
+allowed-tools: Read Grep Glob Bash Edit Write TaskCreate TaskList TaskUpdate
 ---
 
 # Exploit development

package/plugins/core/reverse-skills/skills/forensics-threat-hunting/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: forensics-threat-hunting
 description: Perform disk, memory, packet, and log forensics; build timelines and IOCs; create YARA/Sigma detections; and conduct evidence-driven threat hunts.
-allowed-tools: Read Grep Glob Bash Edit Write
+allowed-tools: Read Grep Glob Bash Edit Write TaskCreate TaskList TaskUpdate
 ---
 
 # Forensics and threat hunting

package/plugins/core/reverse-skills/skills/llm-agent-security/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: llm-agent-security
 description: Test LLM and agent systems for prompt injection, tool abuse, memory poisoning, data leakage, unsafe autonomy, identity failures, and model or skill supply-chain risks.
-allowed-tools: Read Grep Glob Bash Edit Write WebFetch WebSearch
+allowed-tools: Read Grep Glob Bash Edit Write WebFetch WebSearch TaskCreate TaskList TaskUpdate
 ---
 
 # LLM and agent security

package/plugins/core/reverse-skills/skills/malware-edr-research/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: malware-edr-research
 description: Analyze, modify, build, and test malware-like samples, implants, anti-analysis behavior, and EDR/AV telemetry or evasion in authorized research environments.
-allowed-tools: Read Grep Glob Bash Edit Write
+allowed-tools: Read Grep Glob Bash Edit Write TaskCreate TaskList TaskUpdate
 ---
 
 # Malware and EDR research

package/plugins/core/reverse-skills/skills/platform-radio-security/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: platform-radio-security
 description: Research mobile apps, firmware, embedded hardware, OT/ICS systems, Wi-Fi, wireless, and SDR/radio targets in owned or explicitly authorized labs.
-allowed-tools: Read Grep Glob Bash Edit Write
+allowed-tools: Read Grep Glob Bash Edit Write TaskCreate TaskList TaskUpdate
 ---
 
 # Platform, device, and radio security

package/plugins/core/reverse-skills/skills/research-evidence/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: research-evidence
 description: Establish scope, review case integrity, preserve evidence-to-finding traceability, generate security diagrams, and produce reproducible technical reports.
-allowed-tools: Read Grep Glob Bash Edit Write
+allowed-tools: Read Grep Glob Bash Edit Write TaskCreate TaskList TaskUpdate
 ---
 
 # Research evidence and reporting

package/plugins/core/reverse-skills/skills/reverse-skill-router/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: reverse-skill-router
 description: Route an authorized reverse engineering or red-team research task to the correct UR specialist skill. Use when a task spans domains or its best entrypoint is unclear.
-allowed-tools: Read Grep Glob Bash
+allowed-tools: Read Grep Glob Bash Skill TaskCreate TaskList TaskUpdate
 ---
 
 # Reverse Skills router