@yemi33/minions 0.1.1102 → 0.1.1104

package/CHANGELOG.md

@@ -1,11 +1,12 @@
 # Changelog
 
-## 0.1.1102 (2026-04-18)
+## 0.1.1104 (2026-04-18)
 
 ### Features
 - seed realActivityMap at spawn time, stamp pid in live-output (#1200)
 
 ### Fixes
+- preserve VERDICT marker by tail-slicing agent output (#1234) (#1270)
 - PRD info cache staleness and aggregate PR bleed-through (#1222)
 -  avoid no-op work item writes
 -  resilient claude binary resolution + surface spawn errors
@@ -18,6 +19,40 @@
 - stamp live-output.log stub before spawn (#1198)
 -  harden settings save and migrate pr poll config
 
+### Other
+- refactor: clarify settings page section structure and PR polling dependencies
+- refactor: extract _probeClaudePackage helper, use shared.log for spawn errors
+- Make work item descriptions scrollable
+- Use PAT for publish merges
+- test(queries): add unit tests for invalidateDispatchCache/getInbox/getAgentCharter (#1214)
+- test(shared): add unit tests for truncateTextBytes/tailTextBytes/execSilent/trackReviewMetric/parseCanonicalPrId (#1215)
+- Fix publish workflow merge
+- chore: raise default meeting round timeout
+- Harden prompt context handling
+- Harden loop watch conversion
+- Add watches sidebar activity badge
+- test(cli): add unit tests for handleCommand, start, stop, kill, spawn (#1191)
+- chore: untrack pipeline files — local config only
+- restore: recover daily-arch-improvement and weekly-dead-code-cleanup pipelines
+- Harden CC stream resilience
+
+## 0.1.1101 (2026-04-18)
+
+### Features
+- seed realActivityMap at spawn time, stamp pid in live-output (#1200)
+
+### Fixes
+-  avoid no-op work item writes
+-  resilient claude binary resolution + surface spawn errors
+-  resolve native claude.exe from npm wrapper on Windows
+- cap temp-agent creation at maxConcurrent per tick (#1219)
+-  reassign pending items from unspawned temp agents to idle named agents (#1204) (#1212)
+- guard undefined agent in pending dispatch loop (closes #1206) (#1210)
+-  improve fallback meeting conclusion
+-  remove command center chevron
+- stamp live-output.log stub before spawn (#1198)
+-  harden settings save and migrate pr poll config
+
 ### Other
 - refactor: extract _probeClaudePackage helper, use shared.log for spawn errors
 - Make work item descriptions scrollable

package/dashboard/js/settings.js

@@ -47,22 +47,33 @@ async function openSettings() {
       settingsField('Restart Grace Period', 'set-restartGracePeriod', e.restartGracePeriod || 1200000, 'ms', 'Grace period before orphan detection on restart') +
       settingsField('Meeting Round Timeout', 'set-meetingRoundTimeout', e.meetingRoundTimeout || 900000, 'ms', 'Auto-advance meeting round after this') +
     '</div>' +
+    '<h3 style="font-size:13px;color:var(--blue);margin-bottom:8px">Automation</h3>' +
     '<div style="display:flex;flex-direction:column;gap:6px;margin-bottom:16px">' +
       settingsToggle('Auto-approve Plans', 'set-autoApprovePlans', !!e.autoApprovePlans, 'PRDs are approved automatically without human review') +
       settingsToggle('Eval Loop', 'set-evalLoop', e.evalLoop !== false, 'Auto-review implementations and iterate fix cycles until pass') +
       settingsToggle('Auto-decompose', 'set-autoDecompose', e.autoDecompose !== false, 'Large implement items are auto-split into sub-tasks') +
       settingsToggle('Allow Temp Agents', 'set-allowTempAgents', !!e.allowTempAgents, 'Spawn ephemeral agents when all permanent agents are busy') +
       settingsToggle('Auto-archive Plans', 'set-autoArchive', !!e.autoArchive, 'Automatically archive plans after verify completes (off = manual archive via dashboard)') +
-      settingsToggle('Auto-fix Builds', 'set-autoFixBuilds', e.autoFixBuilds !== false, 'Auto-dispatch fix agents when a PR build fails') +
-      settingsToggle('Auto-fix Conflicts', 'set-autoFixConflicts', e.autoFixConflicts !== false, 'Auto-dispatch fix agents when a PR has merge conflicts') +
       settingsToggle('Auto-complete PRs', 'set-autoCompletePrs', !!e.autoCompletePrs, 'Auto-merge PRs when builds pass and review is approved (opt-in)') +
-      settingsToggle('ADO Polling', 'set-adoPollEnabled', e.adoPollEnabled !== false, 'Poll ADO PR status and comments each tick (reconciliation always runs)') +
-      settingsToggle('GitHub Polling', 'set-ghPollEnabled', e.ghPollEnabled !== false, 'Poll GitHub PR status and comments each tick (reconciliation always runs)') +
     '</div>' +
-    '<div style="display:grid;grid-template-columns:1fr 1fr;gap:8px;margin-bottom:16px">' +
-      settingsField('PR Status Poll Frequency', 'set-prPollStatusEvery', e.prPollStatusEvery ?? e.adoPollStatusEvery ?? 12, 'ticks', 'Poll PR build/review/merge status every N ticks for both ADO and GitHub (~12 min at default tick rate)') +
-      settingsField('PR Comments Poll Frequency', 'set-prPollCommentsEvery', e.prPollCommentsEvery ?? e.adoPollCommentsEvery ?? 12, 'ticks', 'Poll PR human comments every N ticks for both ADO and GitHub (~12 min at default tick rate)') +
+
+    '<h3 style="font-size:13px;color:var(--blue);margin-bottom:8px">PR Polling &amp; Dispatch Gates</h3>' +
+    '<div style="border:1px solid var(--border);border-radius:6px;padding:10px 12px;margin-bottom:16px">' +
+      '<div style="display:flex;flex-direction:column;gap:6px;margin-bottom:10px">' +
+        settingsToggle('ADO Polling', 'set-adoPollEnabled', e.adoPollEnabled !== false, 'Keeps ADO PR build results, votes, and comments fresh each tick — the two fix gates below are silently inert when this is off') +
+        '<div style="margin-left:20px;padding-left:10px;border-left:2px solid var(--border);display:flex;flex-direction:column;gap:4px">' +
+          settingsToggle('Auto-fix Builds', 'set-autoFixBuilds', e.autoFixBuilds !== false, 'Dispatch gate: auto-fix agent when build fails (downstream of ADO Polling)') +
+          settingsToggle('Auto-fix Conflicts', 'set-autoFixConflicts', e.autoFixConflicts !== false, 'Dispatch gate: auto-fix agent when merge conflict detected (downstream of ADO Polling)') +
+        '</div>' +
+      '</div>' +
+      settingsToggle('GitHub Polling', 'set-ghPollEnabled', e.ghPollEnabled !== false, 'Keeps GitHub PR build results, votes, and comments fresh each tick (reconciliation always runs regardless)') +
+      '<div style="display:grid;grid-template-columns:1fr 1fr;gap:8px;margin-top:10px">' +
+        settingsField('Status Poll Frequency', 'set-prPollStatusEvery', e.prPollStatusEvery ?? e.adoPollStatusEvery ?? 12, 'ticks', 'Poll PR build/review/merge status every N ticks for both ADO and GitHub (~12 min at default tick rate)') +
+        settingsField('Comments Poll Frequency', 'set-prPollCommentsEvery', e.prPollCommentsEvery ?? e.adoPollCommentsEvery ?? 12, 'ticks', 'Poll PR human comments every N ticks for both ADO and GitHub (~12 min at default tick rate)') +
+      '</div>' +
     '</div>' +
+
+    '<h3 style="font-size:13px;color:var(--blue);margin-bottom:8px">Limits &amp; Thresholds</h3>' +
     '<div style="display:grid;grid-template-columns:1fr 1fr;gap:8px;margin-bottom:16px">' +
       settingsField('Eval Max Iterations', 'set-evalMaxIterations', e.evalMaxIterations || 3, '', 'Max review→fix cycles before escalating (1-10)') +
       settingsField('Eval Max Cost', 'set-evalMaxCost', e.evalMaxCost === null || e.evalMaxCost === undefined ? '' : e.evalMaxCost, '$', 'USD ceiling per work item across all eval iterations (blank = no limit)') +
@@ -78,7 +89,7 @@ async function openSettings() {
       return '<div style="border:1px solid var(--border);border-radius:6px;padding:10px 12px">' +
         '<div style="font-size:12px;font-weight:600;margin-bottom:8px">' + escHtml(p.name) + '</div>' +
         '<div style="display:flex;flex-direction:column;gap:6px">' +
-        settingsToggle('Discover from PRs', 'set-ws-prs-' + p.name, p.workSources.pullRequests.enabled, 'Auto-discover work from open pull requests') +
+        settingsToggle('Discover from PRs', 'set-ws-prs-' + p.name, p.workSources.pullRequests.enabled, 'Discovery gate: scan repo for open PRs and surface them as review tasks. Independent of ADO/GitHub polling — does not affect already-tracked PRs.') +
         settingsToggle('Discover from Work Items', 'set-ws-wi-' + p.name, p.workSources.workItems.enabled, 'Auto-discover work from ADO/GitHub work items') +
         '</div></div>';
     }).join('') +

package/engine/llm.js

@@ -149,7 +149,9 @@ function _createStreamAccumulator({
     if (obj.session_id) sessionId = obj.session_id;
     if (obj.type === 'result') {
       if (typeof obj.result === 'string') {
-        text = maxTextLength ? obj.result.slice(0, maxTextLength) : obj.result;
+        // Tail-slice: VERDICTs, completion blocks, and PR URLs live at the END
+        // of agent output. Head-slicing dropped them (#1234).
+        text = maxTextLength ? obj.result.slice(-maxTextLength) : obj.result;
       }
       if (obj.total_cost_usd || obj.usage) {
         usage = {
@@ -166,7 +168,8 @@ function _createStreamAccumulator({
     if (obj.type === 'assistant' && Array.isArray(obj.message?.content)) {
       for (const block of obj.message.content) {
         if (block?.type === 'text' && block.text) {
-          text = maxTextLength ? block.text.slice(0, maxTextLength) : block.text;
+          // Tail-slice for consistency with the result branch (see #1234).
+          text = maxTextLength ? block.text.slice(-maxTextLength) : block.text;
           if (onChunk && block.text !== lastTextSent) {
             lastTextSent = block.text;
             onChunk(block.text);

package/engine/shared.js

@@ -608,7 +608,11 @@ function parseStreamJsonOutput(raw, { maxTextLength = 0 } = {}) {
 
   function extractResult(obj) {
     if (obj.type !== 'result') return false;
-    if (obj.result) text = maxTextLength ? obj.result.slice(0, maxTextLength) : obj.result;
+    // Slice from the tail, not the head — review VERDICTs, structured completion
+    // blocks, PR URLs, and agent conclusions all appear at the END of the output.
+    // Head-slicing truncated VERDICTs and caused review work items to be
+    // re-dispatched up to maxRetries times despite successful completion (#1234).
+    if (obj.result) text = maxTextLength ? obj.result.slice(-maxTextLength) : obj.result;
     if (obj.session_id) sessionId = obj.session_id;
     if (obj.total_cost_usd || obj.usage) {
       usage = {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yemi33/minions",
-  "version": "0.1.1102",
+  "version": "0.1.1104",
   "description": "Multi-agent AI dev team that runs from ~/.minions/ — five autonomous agents share a single engine, dashboard, and knowledge base",
   "bin": {
     "minions": "bin/minions.js"