@yemi33/minions 0.1.227 → 0.1.229

package/CHANGELOG.md

@@ -1,8 +1,22 @@
 # Changelog
 
-## 0.1.227 (2026-04-03)
+## 0.1.229 (2026-04-03)
+
+### Fixes
+- consolidate temp agent metrics into one row
+
+## 0.1.228 (2026-04-03)
+
+### Features
+-  scan for projects UI — modal with checkbox multi-select
+
+### Fixes
+-  CC system prompt now defaults to delegating work to agents
+
+## 0.1.226 (2026-04-03)
 
 ### Features
+- add evaluate type to routing and playbooks (#70)
 - fix cooldown context accumulation bloat (#71)
 - wire dead test and extract progress bar helper (#64)
 

package/dashboard/js/render-other.js

@@ -16,7 +16,8 @@ function renderProjects(projects) {
       (p.description ? '<span style="color:var(--muted);font-weight:400;margin-left:6px;font-size:10px">' + escHtml(p.description.slice(0, 60)) + (p.description.length > 60 ? '...' : '') + '</span>' : '') +
     '</span>'
   ).join('') +
-  '<span onclick="addProject()" style="background:var(--surface2);border:1px solid var(--border);border-radius:4px;padding:3px 10px;color:var(--muted);font-weight:500;cursor:pointer;border-style:dashed">+ Add</span>';
+  '<span onclick="addProject()" style="background:var(--surface2);border:1px solid var(--border);border-radius:4px;padding:3px 10px;color:var(--muted);font-weight:500;cursor:pointer;border-style:dashed">+ Add</span>' +
+  '<span onclick="openScanProjectsModal()" style="background:var(--surface2);border:1px solid var(--border);border-radius:4px;padding:3px 10px;color:var(--blue);font-weight:500;cursor:pointer;border-style:dashed;font-size:10px">Scan</span>';
 
 }
 
@@ -47,8 +48,26 @@ function renderMetrics(metrics) {
     renderContextPressure(metrics);
     return;
   }
+  // Consolidate temp-* agents into one row
+  var permanent = agents.filter(function(a) { return !a[0].startsWith('temp-'); });
+  var temps = agents.filter(function(a) { return a[0].startsWith('temp-'); });
+  var rows = permanent.slice();
+  if (temps.length > 0) {
+    var merged = { tasksCompleted: 0, tasksErrored: 0, prsCreated: 0, prsApproved: 0, prsRejected: 0, reviewsDone: 0 };
+    for (var t = 0; t < temps.length; t++) {
+      var tm = temps[t][1];
+      merged.tasksCompleted += tm.tasksCompleted || 0;
+      merged.tasksErrored += tm.tasksErrored || 0;
+      merged.prsCreated += tm.prsCreated || 0;
+      merged.prsApproved += tm.prsApproved || 0;
+      merged.prsRejected += tm.prsRejected || 0;
+      merged.reviewsDone += tm.reviewsDone || 0;
+    }
+    rows.push(['Temp Agents (' + temps.length + ')', merged]);
+  }
+
   let html = '<table class="pr-table"><thead><tr><th>Agent</th><th>Done</th><th>Errors</th><th>PRs</th><th>Approved</th><th>Rejected</th><th>Rate</th><th>Reviews</th></tr></thead><tbody>';
-  for (const [id, m] of agents) {
+  for (const [id, m] of rows) {
     const rate = m.prsCreated > 0 ? Math.round((m.prsApproved / m.prsCreated) * 100) + '%' : '-';
     const rateColor = m.prsCreated > 0 ? (m.prsApproved / m.prsCreated >= 0.7 ? 'var(--green)' : 'var(--red)') : 'var(--muted)';
     html += '<tr>' +
@@ -140,8 +159,25 @@ function renderTokenUsage(metrics) {
     html += '</div>';
   }
 
-  // Per-agent token table
-  const agentsWithUsage = agents.filter(([, m]) => (m.totalCostUsd || 0) > 0);
+  // Per-agent token table (consolidate temp agents)
+  var permAgents = agents.filter(function(a) { return !a[0].startsWith('temp-'); });
+  var tempAgents = agents.filter(function(a) { return a[0].startsWith('temp-'); });
+  var tokenRows = permAgents.slice();
+  if (tempAgents.length > 0) {
+    var tempMerged = { totalCostUsd: 0, totalInputTokens: 0, totalOutputTokens: 0, totalCacheRead: 0, tasksCompleted: 0, tasksErrored: 0, model: '' };
+    for (var ti = 0; ti < tempAgents.length; ti++) {
+      var tm = tempAgents[ti][1];
+      tempMerged.totalCostUsd += tm.totalCostUsd || 0;
+      tempMerged.totalInputTokens += tm.totalInputTokens || 0;
+      tempMerged.totalOutputTokens += tm.totalOutputTokens || 0;
+      tempMerged.totalCacheRead += tm.totalCacheRead || 0;
+      tempMerged.tasksCompleted += tm.tasksCompleted || 0;
+      tempMerged.tasksErrored += tm.tasksErrored || 0;
+      if (!tempMerged.model && tm.model) tempMerged.model = tm.model;
+    }
+    tokenRows.push(['Temp Agents (' + tempAgents.length + ')', tempMerged]);
+  }
+  var agentsWithUsage = tokenRows.filter(function(a) { return (a[1].totalCostUsd || 0) > 0; });
   if (agentsWithUsage.length > 0) {
     html += '<div style="font-size:10px;color:var(--muted);margin:12px 0 4px;font-weight:600;text-transform:uppercase;letter-spacing:0.5px">Agent Usage</div>';
     html += '<table class="token-agent-table"><thead><tr><th>Agent</th><th>Model</th><th>Cost</th><th>Input</th><th>Output</th><th>Cache</th><th>$/task</th></tr></thead><tbody>';
@@ -206,4 +242,99 @@ function renderContextPressure(metrics) {
   el.innerHTML = html;
 }
 
-window.MinionsOther = { renderProjects, renderMcpServers, renderMetrics, renderTokenUsage, renderContextPressure };
+async function openScanProjectsModal() {
+  document.getElementById('modal-title').textContent = 'Scan for Projects';
+  document.getElementById('modal-body').innerHTML =
+    '<div style="display:flex;flex-direction:column;gap:12px">' +
+      '<div style="display:flex;gap:8px;align-items:flex-end">' +
+        '<label style="flex:1;color:var(--text);font-size:var(--text-md)">Directory to scan' +
+          '<input id="scan-path" value="' + escHtml((typeof os !== 'undefined' ? os.homedir() : '~') || '~') + '" style="display:block;width:100%;margin-top:4px;padding:6px 8px;background:var(--bg);border:1px solid var(--border);border-radius:var(--radius-sm);color:var(--text);font-size:var(--text-md)">' +
+        '</label>' +
+        '<label style="width:60px;color:var(--text);font-size:var(--text-md)">Depth' +
+          '<input id="scan-depth" type="number" value="3" min="1" max="6" style="display:block;width:100%;margin-top:4px;padding:6px 8px;background:var(--bg);border:1px solid var(--border);border-radius:var(--radius-sm);color:var(--text);font-size:var(--text-md)">' +
+        '</label>' +
+        '<button onclick="_runProjectScan()" style="padding:6px 16px;background:var(--blue);color:#fff;border:none;border-radius:var(--radius-sm);cursor:pointer;white-space:nowrap">Scan</button>' +
+      '</div>' +
+      '<div id="scan-results" style="color:var(--muted);font-size:12px">Click Scan to find git repos in the directory.</div>' +
+    '</div>';
+  document.getElementById('modal-body').style.whiteSpace = 'normal';
+  document.getElementById('modal-body').style.fontFamily = "'Segoe UI', system-ui, sans-serif";
+  document.getElementById('modal').classList.add('open');
+}
+
+async function _runProjectScan() {
+  var scanPath = document.getElementById('scan-path')?.value?.trim();
+  var depth = document.getElementById('scan-depth')?.value || '3';
+  var resultsEl = document.getElementById('scan-results');
+  if (!scanPath) { alert('Enter a path'); return; }
+  resultsEl.innerHTML = '<span style="color:var(--blue)">Scanning...</span>';
+
+  try {
+    var res = await fetch('/api/projects/scan', {
+      method: 'POST', headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({ path: scanPath, depth: Number(depth) })
+    });
+    var data = await res.json();
+    if (!res.ok) { resultsEl.innerHTML = '<span style="color:var(--red)">Error: ' + escHtml(data.error) + '</span>'; return; }
+    var repos = data.repos || [];
+    if (repos.length === 0) { resultsEl.innerHTML = '<span style="color:var(--muted)">No git repos found in ' + escHtml(scanPath) + '</span>'; return; }
+
+    var html = '<div style="margin-bottom:8px;font-size:11px;color:var(--muted)">' + repos.length + ' repos found — select to add:</div>';
+    html += '<div style="max-height:400px;overflow-y:auto;display:flex;flex-direction:column;gap:4px">';
+    repos.forEach(function(r, i) {
+      var linked = r.linked;
+      var hostBadge = '<span style="font-size:9px;padding:1px 5px;border-radius:3px;background:' +
+        (r.host === 'GitHub' ? 'rgba(88,166,255,0.15);color:var(--blue)' : r.host === 'ADO' ? 'rgba(188,140,255,0.15);color:var(--purple)' : 'var(--surface2);color:var(--muted)') +
+        '">' + escHtml(r.host) + '</span>';
+      html += '<label style="display:flex;align-items:center;gap:8px;padding:6px 10px;background:var(--surface2);border:1px solid var(--border);border-radius:4px;cursor:' + (linked ? 'default' : 'pointer') + ';opacity:' + (linked ? '0.5' : '1') + '">' +
+        '<input type="checkbox" data-scan-idx="' + i + '" ' + (linked ? 'disabled checked' : '') + ' style="accent-color:var(--blue);width:16px;height:16px">' +
+        '<div style="flex:1;min-width:0">' +
+          '<div style="font-weight:600;font-size:12px">' + escHtml(r.name) + ' ' + hostBadge + (linked ? ' <span style="font-size:9px;color:var(--green)">linked</span>' : '') + '</div>' +
+          '<div style="font-size:10px;color:var(--muted);overflow:hidden;text-overflow:ellipsis;white-space:nowrap">' + escHtml(r.path) + '</div>' +
+          (r.description ? '<div style="font-size:10px;color:var(--muted)">' + escHtml(r.description) + '</div>' : '') +
+        '</div>' +
+      '</label>';
+    });
+    html += '</div>';
+    html += '<div style="display:flex;justify-content:space-between;align-items:center;margin-top:8px">' +
+      '<div style="display:flex;gap:8px">' +
+        '<span onclick="_scanSelectAll()" style="font-size:10px;color:var(--blue);cursor:pointer;text-decoration:underline">Select all</span>' +
+        '<span onclick="_scanSelectNone()" style="font-size:10px;color:var(--muted);cursor:pointer;text-decoration:underline">Clear</span>' +
+      '</div>' +
+      '<button onclick="_addSelectedProjects()" style="padding:6px 16px;background:var(--green);color:#fff;border:none;border-radius:var(--radius-sm);cursor:pointer">Add Selected</button>' +
+    '</div>';
+    resultsEl.innerHTML = html;
+    window._scanRepos = repos;
+  } catch (e) { resultsEl.innerHTML = '<span style="color:var(--red)">Error: ' + escHtml(e.message) + '</span>'; }
+}
+
+function _scanSelectAll() {
+  document.querySelectorAll('[data-scan-idx]').forEach(function(cb) { if (!cb.disabled) cb.checked = true; });
+}
+function _scanSelectNone() {
+  document.querySelectorAll('[data-scan-idx]').forEach(function(cb) { if (!cb.disabled) cb.checked = false; });
+}
+
+async function _addSelectedProjects() {
+  var checkboxes = document.querySelectorAll('[data-scan-idx]:checked:not(:disabled)');
+  if (checkboxes.length === 0) { alert('Select at least one repo'); return; }
+  var repos = window._scanRepos || [];
+  var added = 0;
+  for (var cb of checkboxes) {
+    var repo = repos[parseInt(cb.dataset.scanIdx)];
+    if (!repo) continue;
+    try {
+      var res = await fetch('/api/projects/add', {
+        method: 'POST', headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({ path: repo.path })
+      });
+      if (res.ok) { added++; cb.disabled = true; cb.closest('label').style.opacity = '0.5'; }
+    } catch { /* continue with next */ }
+  }
+  if (added > 0) {
+    showToast('cmd-toast', added + ' project(s) added', true);
+    refresh();
+  }
+}
+
+window.MinionsOther = { renderProjects, renderMcpServers, renderMetrics, renderTokenUsage, renderContextPressure, openScanProjectsModal };

package/dashboard/layout.html

@@ -52,8 +52,8 @@
   <div style="font-size:18px;margin-bottom:6px;text-align:center">Getting Started</div>
   <div style="color:var(--muted);font-size:13px;margin-bottom:12px;text-align:center">No projects linked yet. Link a project so agents know what code to work on.</div>
   <div style="display:flex;flex-direction:column;gap:8px;align-items:center">
-    <code style="background:var(--bg);padding:6px 16px;border-radius:4px;font-size:13px;color:var(--blue);border:1px solid var(--border)">minions add &lt;path-to-your-repo&gt;</code>
-    <div style="font-size:11px;color:var(--muted)">Then try: <code style="background:var(--bg);padding:2px 6px;border-radius:3px;font-size:11px">minions work "Explore the codebase"</code> or use the Command Center above</div>
+    <button onclick="openScanProjectsModal()" style="padding:8px 20px;background:var(--blue);color:#fff;border:none;border-radius:var(--radius-sm);cursor:pointer;font-size:13px;font-weight:600">Scan for Projects</button>
+    <div style="font-size:11px;color:var(--muted)">Or via CLI: <code style="background:var(--bg);padding:2px 6px;border-radius:3px;font-size:11px">minions add &lt;path-to-your-repo&gt;</code></div>
   </div>
 </div>
 

package/dashboard.js

@@ -340,19 +340,26 @@ ${MINIONS_DIR}/
 
 Projects are configured in \`config.json\` under \`projects[]\`. Per-project state lives centrally in \`${MINIONS_DIR}/projects/{name}/\` — NOT inside project repos. There are no \`.minions/\` folders inside project repos.
 
-## Direct Execution
+## Default: Delegate to Agents
 
-You have Bash, Write, Edit, and all standard tools. Use them directly when the task is straightforward:
-- **Build & run projects** — \`cd <project> && npm install && npm run dev\`
-- **Inspect code** — read files, grep, explore
-- **Edit project files** — fix configs, update docs, tweak settings
-- **Git operations** — fetch, checkout, merge, diff (but do NOT push without the user confirming)
-- **Start dev servers** — for long-running servers, use detached processes so they survive after you finish
+Your primary job is to **orchestrate**, not implement. For most requests, dispatch work to agents rather than doing it yourself. Agents have full Claude Code sessions with project context, worktrees, and MCP tools — they are better equipped for real work.
 
-**When to do it yourself vs delegate to an agent:**
-- Quick, one-shot tasks (build, read, check, install, start a server) → do it yourself
-- Complex multi-file code changes, PR creation, code review → dispatch to an agent
-- Anything that needs deep codebase knowledge or iterative coding → dispatch to an agent
+**Delegate to agents (default):**
+- Code changes of any size — even "small" fixes
+- Bug fixes, feature implementation, refactoring
+- PR creation and code review
+- Testing and verification
+- Codebase exploration and architecture analysis
+- Plan creation and execution
+
+**Do it yourself (only these):**
+- Reading a specific file or status to answer a question
+- Quick lookups (check a config value, find a file, read an error log)
+- Creating notes, editing plans, updating routing/charters
+- Starting a dev server or running a build command the user asked for
+- Git operations the user explicitly asked YOU to do
+
+When in doubt, **dispatch an agent**. The user is talking to you because they want work delegated, not because they want you to be the one coding.
 
 ## Minions Actions (Delegation)
 
@@ -2904,6 +2911,52 @@ What would you like to discuss or change? When you're happy, say "approve" and I
     } catch (e) { return jsonReply(res, 400, { error: e.message }); }
   }
 
+  async function handleProjectsScan(req, res) {
+    try {
+      const body = await readBody(req);
+      const scanRoot = path.resolve(body.path || os.homedir());
+      const maxDepth = Math.min(Number(body.depth) || 3, 6);
+      if (!fs.existsSync(scanRoot)) return jsonReply(res, 400, { error: 'path does not exist' });
+
+      // Find git repos recursively (same logic as minions.js findGitRepos)
+      const skipDirs = new Set(['node_modules', '.git', '.hg', 'AppData', '$Recycle.Bin', 'Windows',
+        'Program Files', 'Program Files (x86)', '.cache', '.npm', '.yarn', '.nuget', 'worktrees', '.minions', '.squad']);
+      const repos = [];
+      function walk(dir, depth) {
+        if (depth > maxDepth) return;
+        let entries;
+        try { entries = fs.readdirSync(dir, { withFileTypes: true }); } catch { return; }
+        for (const e of entries) {
+          if (!e.isDirectory() || (e.name.startsWith('.') && e.name !== '.git') || skipDirs.has(e.name)) continue;
+          if (e.name === '.git') { repos.push(dir); return; }
+          walk(path.join(dir, e.name), depth + 1);
+        }
+      }
+      walk(scanRoot, 0);
+
+      // Enrich each repo with metadata
+      const existingPaths = new Set(PROJECTS.map(p => path.resolve(p.localPath)));
+      const results = repos.map(repoPath => {
+        const result = { path: repoPath.replace(/\\/g, '/'), name: path.basename(repoPath), host: 'git', linked: existingPaths.has(path.resolve(repoPath)) };
+        try {
+          const remoteUrl = require('child_process').execSync('git remote get-url origin', { cwd: repoPath, encoding: 'utf8', timeout: 3000, stdio: ['pipe', 'pipe', 'pipe'] }).trim();
+          const gh = remoteUrl.match(/github\.com[:/]([^/]+)\/([^/.]+)/);
+          const ado = remoteUrl.match(/dev\.azure\.com\/([^/]+)\/([^/]+)\/_git\/([^/\s]+)/) || remoteUrl.match(/([^.]+)\.visualstudio\.com.*?\/([^/]+)\/_git\/([^/\s]+)/);
+          if (gh) { result.host = 'GitHub'; result.org = gh[1]; result.name = gh[2]; }
+          else if (ado) { result.host = 'ADO'; result.org = ado[1]; result.name = ado[3] || ado[2]; }
+        } catch { /* no remote */ }
+        try {
+          const pkg = JSON.parse(fs.readFileSync(path.join(repoPath, 'package.json'), 'utf8'));
+          if (pkg.name) result.name = pkg.name.replace(/@[^/]+\//, '');
+          if (pkg.description) result.description = pkg.description.slice(0, 100);
+        } catch { /* no package.json */ }
+        return result;
+      });
+
+      return jsonReply(res, 200, { repos: results });
+    } catch (e) { return jsonReply(res, 500, { error: e.message }); }
+  }
+
   async function handleCommandCenterNewSession(req, res) {
     ccSession = { sessionId: null, createdAt: null, lastActiveAt: null, turnCount: 0 };
     ccInFlight = false; // Reset concurrency guard so a stuck request doesn't block new sessions
@@ -3444,6 +3497,7 @@ What would you like to discuss or change? When you're happy, say "approve" and I
 
     // Projects
     { method: 'POST', path: '/api/projects/browse', desc: 'Open folder picker dialog, return selected path', handler: handleProjectsBrowse },
+    { method: 'POST', path: '/api/projects/scan', desc: 'Scan a directory for git repos', params: 'path?, depth?', handler: handleProjectsScan },
     { method: 'POST', path: '/api/projects/add', desc: 'Auto-discover and add a project to config', params: 'path, name?', handler: handleProjectsAdd },
 
     // Command Center

package/engine/cleanup.js

@@ -35,20 +35,23 @@ function runCleanup(config, verbose = false) {
   const projects = getProjects(config);
   let cleaned = { tempFiles: 0, liveOutputs: 0, worktrees: 0, zombies: 0 };
 
-  // 1. Clean stale temp prompt/sysprompt files (older than 1 hour)
+  // 1. Clean stale temp prompt/sysprompt files and orphaned safeWrite .tmp.* files (older than 1 hour)
   const oneHourAgo = Date.now() - 3600000;
   try {
     const tmpDir = path.join(ENGINE_DIR, 'tmp');
     const scanDirs = [ENGINE_DIR, ...(fs.existsSync(tmpDir) ? [tmpDir] : [])];
     for (const dir of scanDirs) {
       for (const f of fs.readdirSync(dir)) {
-        if (f.startsWith('prompt-') || f.startsWith('sysprompt-') || f.startsWith('tmp-sysprompt-')) {
+        const isPromptTemp = f.startsWith('prompt-') || f.startsWith('sysprompt-') || f.startsWith('tmp-sysprompt-');
+        const isSafeWriteTemp = /\.tmp\.\d+\.\d+$/.test(f);
+        if (isPromptTemp || isSafeWriteTemp) {
           const fp = path.join(dir, f);
           try {
             const stat = fs.statSync(fp);
             if (stat.mtimeMs < oneHourAgo) {
               fs.unlinkSync(fp);
               cleaned.tempFiles++;
+              if (isSafeWriteTemp) log('info', `Cleaned orphaned temp file: ${f}`);
             }
           } catch { /* cleanup */ }
         }

package/engine/lifecycle.js

@@ -630,7 +630,22 @@ function syncPrsFromOutput(output, agentId, meta, config) {
   }
 
   for (const [name, entry] of dirtyTargets) {
-    shared.safeWrite(entry.prPath, entry.prs);
+    mutateJsonFileLocked(entry.prPath, (existingPrs) => {
+      // Merge new PRs into existing data to avoid losing concurrent writes
+      for (const pr of entry.prs) {
+        const idx = existingPrs.findIndex(p => p.id === pr.id);
+        if (idx >= 0) {
+          // Backfill prdItems if needed
+          if (pr.prdItems?.length) {
+            const merged = new Set([...(existingPrs[idx].prdItems || []), ...pr.prdItems]);
+            existingPrs[idx].prdItems = [...merged];
+          }
+        } else {
+          existingPrs.push(pr);
+        }
+      }
+      return existingPrs;
+    }, { defaultValue: [] });
     log('info', `Synced PR(s) from ${agentName}'s output to ${name === '_central' ? 'central' : name}/pull-requests.json`);
   }
   return added;
@@ -668,11 +683,12 @@ function updatePrAfterReview(agentId, pr, project) {
   const authorAgentId = (pr.agent || '').toLowerCase();
   if (authorAgentId && config.agents?.[authorAgentId]) {
     const metricsPath = path.join(ENGINE_DIR, 'metrics.json');
-    const metrics = safeJson(metricsPath) || {};
-    if (!metrics[authorAgentId]) metrics[authorAgentId] = { tasksCompleted:0, tasksErrored:0, prsCreated:0, prsApproved:0, prsRejected:0, reviewsDone:0, lastTask:null, lastCompleted:null };
-    if (!metrics[agentId]) metrics[agentId] = { tasksCompleted:0, tasksErrored:0, prsCreated:0, prsApproved:0, prsRejected:0, reviewsDone:0, lastTask:null, lastCompleted:null };
-    metrics[agentId].reviewsDone = (metrics[agentId].reviewsDone || 0) + 1;
-    shared.safeWrite(metricsPath, metrics);
+    mutateJsonFileLocked(metricsPath, (metrics) => {
+      if (!metrics[authorAgentId]) metrics[authorAgentId] = { tasksCompleted:0, tasksErrored:0, prsCreated:0, prsApproved:0, prsRejected:0, reviewsDone:0, lastTask:null, lastCompleted:null };
+      if (!metrics[agentId]) metrics[agentId] = { tasksCompleted:0, tasksErrored:0, prsCreated:0, prsApproved:0, prsRejected:0, reviewsDone:0, lastTask:null, lastCompleted:null };
+      metrics[agentId].reviewsDone = (metrics[agentId].reviewsDone || 0) + 1;
+      return metrics;
+    });
   }
 
   shared.safeWrite(project ? shared.projectPrPath(project) : path.join(path.resolve(MINIONS_DIR, '..'), '.minions', 'pull-requests.json'), prs);
@@ -747,12 +763,16 @@ async function handlePostMerge(pr, project, config, newStatus) {
       const planFiles = fs.readdirSync(prdDir).filter(f => f.endsWith('.json'));
       let updated = 0;
       for (const pf of planFiles) {
-        const plan = safeJson(path.join(prdDir, pf));
+        const prdPath = path.join(prdDir, pf);
+        const plan = safeJson(prdPath);
         if (!plan?.missing_features) continue;
         const feature = plan.missing_features.find(f => f.id === mergedItemId);
         if (feature && feature.status !== 'implemented') {
-          feature.status = 'implemented';
-          shared.safeWrite(path.join(prdDir, pf), plan);
+          mutateJsonFileLocked(prdPath, (data) => {
+            const feat = data.missing_features?.find(f => f.id === mergedItemId);
+            if (feat && feat.status !== 'implemented') feat.status = 'implemented';
+            return data;
+          });
           updated++;
         }
       }
@@ -764,17 +784,19 @@ async function handlePostMerge(pr, project, config, newStatus) {
     for (const p of shared.getProjects(config)) wiPaths.push(shared.projectWorkItemsPath(p));
     for (const wiPath of wiPaths) {
       try {
-        const items = safeJson(wiPath);
-        if (!items) continue;
-        const item = items.find(i => i.id === mergedItemId);
-        if (item && item.status !== 'done') {
-          log('info', `Post-merge: marking work item ${mergedItemId} as done (was ${item.status}) for ${pr.id}`);
-          item.status = 'done';
-          item.completedAt = ts();
-          item._mergedVia = pr.id;
-          shared.safeWrite(wiPath, items);
-          break;
-        }
+        let found = false;
+        mutateJsonFileLocked(wiPath, (items) => {
+          const item = items.find(i => i.id === mergedItemId);
+          if (item && item.status !== 'done') {
+            log('info', `Post-merge: marking work item ${mergedItemId} as done (was ${item.status}) for ${pr.id}`);
+            item.status = 'done';
+            item.completedAt = ts();
+            item._mergedVia = pr.id;
+            found = true;
+          }
+          return items;
+        }, { defaultValue: [] });
+        if (found) break;
       } catch (err) { log('warn', `Post-merge work item update: ${err.message}`); }
     }
   }
@@ -782,10 +804,11 @@ async function handlePostMerge(pr, project, config, newStatus) {
   const agentId = (pr.agent || '').toLowerCase();
   if (agentId && config.agents?.[agentId]) {
     const metricsPath = path.join(ENGINE_DIR, 'metrics.json');
-    const metrics = safeJson(metricsPath) || {};
-    if (!metrics[agentId]) metrics[agentId] = { tasksCompleted:0, tasksErrored:0, prsCreated:0, prsApproved:0, prsRejected:0, prsMerged:0, reviewsDone:0, lastTask:null, lastCompleted:null };
-    metrics[agentId].prsMerged = (metrics[agentId].prsMerged || 0) + 1;
-    shared.safeWrite(metricsPath, metrics);
+    mutateJsonFileLocked(metricsPath, (metrics) => {
+      if (!metrics[agentId]) metrics[agentId] = { tasksCompleted:0, tasksErrored:0, prsCreated:0, prsApproved:0, prsRejected:0, prsMerged:0, reviewsDone:0, lastTask:null, lastCompleted:null };
+      metrics[agentId].prsMerged = (metrics[agentId].prsMerged || 0) + 1;
+      return metrics;
+    });
   }
 
   const teamsUrl = process.env.TEAMS_PLAN_FLOW_URL;
@@ -1229,14 +1252,15 @@ function runPostCompletionHooks(dispatchItem, agentId, code, stdout, config) {
       try {
         const wiPath = resolveWiPath(meta);
         if (wiPath) {
-          const items = safeJson(wiPath) || [];
-          // For fix items, target the original implement parent; for implement, target self
-          const evalTargetId = type === 'fix' ? meta.item._evalParentId : meta.item.id;
-          // Dedup: skip if an evaluate item already exists for this parent
-          const existing = items.find(i => i._evalParentId === evalTargetId && i.type === 'evaluate' && i.status === 'pending');
-          if (existing) {
-            log('info', `Eval loop: evaluate item ${existing.id} already exists for ${evalTargetId}, skipping`);
-          } else {
+          mutateJsonFileLocked(wiPath, (items) => {
+            // For fix items, target the original implement parent; for implement, target self
+            const evalTargetId = type === 'fix' ? meta.item._evalParentId : meta.item.id;
+            // Dedup: skip if an evaluate item already exists for this parent
+            const existing = items.find(i => i._evalParentId === evalTargetId && i.type === 'evaluate' && i.status === 'pending');
+            if (existing) {
+              log('info', `Eval loop: evaluate item ${existing.id} already exists for ${evalTargetId}, skipping`);
+              return items;
+            }
             const parentItem = items.find(i => i.id === evalTargetId);
             const evalItem = {
               id: 'W-' + shared.uid(),
@@ -1256,9 +1280,9 @@ function runPostCompletionHooks(dispatchItem, agentId, code, stdout, config) {
             // Mark parent as eval-dispatched before writing to prevent duplicates on re-run
             if (parentItem) parentItem._evalDispatched = true;
             items.push(evalItem);
-            shared.safeWrite(wiPath, items);
             log('info', `Eval loop: created ${evalItem.id} for completed ${type} ${meta.item.id} (parent: ${evalTargetId})`);
-          }
+            return items;
+          }, { defaultValue: [] });
         }
       } catch (err) {
         log('warn', `Eval loop dispatch error: ${err.message}`);
@@ -1276,9 +1300,9 @@ function runPostCompletionHooks(dispatchItem, agentId, code, stdout, config) {
       if (verdict && !verdict.pass && evalLoop) {
         const wiPath = resolveWiPath(meta);
         if (wiPath) {
-          const items = safeJson(wiPath) || [];
-          const parent = items.find(i => i.id === meta.item._evalParentId);
-          if (parent) {
+          mutateJsonFileLocked(wiPath, (items) => {
+            const parent = items.find(i => i.id === meta.item._evalParentId);
+            if (!parent) return items;
             const iterations = (parent._evalIterations || 0) + 1;
             parent._evalIterations = iterations;
 
@@ -1286,7 +1310,6 @@ function runPostCompletionHooks(dispatchItem, agentId, code, stdout, config) {
               // Max iterations reached — escalate to human review
               parent.status = 'needs-human-review';
               parent._evalEscalatedAt = ts();
-              shared.safeWrite(wiPath, items);
               log('info', `Eval loop: ${parent.id} reached ${iterations}/${maxIter} iterations — escalated to needs-human-review`);
             } else {
               // Create fix work item with evaluator feedback
@@ -1312,10 +1335,10 @@ function runPostCompletionHooks(dispatchItem, agentId, code, stdout, config) {
               // Parent stays 'done' — fix item picks up from here
               parent.status = 'done';
               items.push(fixItem);
-              shared.safeWrite(wiPath, items);
               log('info', `Eval loop: created fix ${fixItem.id} for failed eval on ${parent.id} (iteration ${iterations}/${maxIter})`);
             }
-          }
+            return items;
+          }, { defaultValue: [] });
         }
       }
     } catch (err) {
@@ -1345,13 +1368,14 @@ function runPostCompletionHooks(dispatchItem, agentId, code, stdout, config) {
           ? path.join(MINIONS_DIR, 'work-items.json')
           : meta.project?.name ? path.join(MINIONS_DIR, 'projects', meta.project.name, 'work-items.json') : null;
         if (wiPath) {
-          const items = safeJson(wiPath) || [];
-          const wi = items.find(i => i.id === meta.item.id);
-          if (wi) {
-            wi._retryCount = retries + 1; wi.status = 'pending'; delete wi.dispatched_at; delete wi.dispatched_to;
-            if (type === 'decompose') delete wi._decomposing; // clear so item can retry decomposition
-            shared.safeWrite(wiPath, items);
-          }
+          mutateJsonFileLocked(wiPath, (items) => {
+            const wi = items.find(i => i.id === meta.item.id);
+            if (wi) {
+              wi._retryCount = retries + 1; wi.status = 'pending'; delete wi.dispatched_at; delete wi.dispatched_to;
+              if (type === 'decompose') delete wi._decomposing; // clear so item can retry decomposition
+            }
+            return items;
+          }, { defaultValue: [] });
         }
       } catch (err) { log('warn', `Retry update: ${err.message}`); }
     } else {
@@ -1364,9 +1388,11 @@ function runPostCompletionHooks(dispatchItem, agentId, code, stdout, config) {
           ? path.join(MINIONS_DIR, 'work-items.json')
           : meta.project?.name ? path.join(MINIONS_DIR, 'projects', meta.project.name, 'work-items.json') : null;
         if (wiPath) {
-          const items = safeJson(wiPath) || [];
-          const wi = items.find(i => i.id === meta.item.id);
-          if (wi) { delete wi._decomposing; shared.safeWrite(wiPath, items); }
+          mutateJsonFileLocked(wiPath, (items) => {
+            const wi = items.find(i => i.id === meta.item.id);
+            if (wi) { delete wi._decomposing; }
+            return items;
+          }, { defaultValue: [] });
         }
       } catch (err) { log('warn', `Decompose cleanup: ${err.message}`); }
     }

package/engine/playbook.js

@@ -481,7 +481,7 @@ function selectPlaybook(workType, item) {
   if (workType === 'review' && !item?._pr && !item?.pr_id) {
     return 'work-item';
   }
-  const typeSpecificPlaybooks = ['explore', 'review', 'test', 'plan-to-prd', 'plan', 'ask', 'verify', 'decompose', 'meeting-investigate', 'meeting-debate', 'meeting-conclude'];
+  const typeSpecificPlaybooks = ['explore', 'review', 'test', 'plan-to-prd', 'plan', 'ask', 'verify', 'decompose', 'evaluate', 'meeting-investigate', 'meeting-debate', 'meeting-conclude'];
   return typeSpecificPlaybooks.includes(workType) ? workType : 'work-item';
 }
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yemi33/minions",
-  "version": "0.1.227",
+  "version": "0.1.229",
   "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"

package/playbooks/evaluate.md

@@ -0,0 +1,149 @@
+# Playbook: Evaluate
+
+You are {{agent_name}}, the {{agent_role}} on the {{project_name}} project.
+TEAM ROOT: {{team_root}}
+
+## Your Task
+
+Evaluate the implementation quality of a completed work item against its acceptance criteria and code quality standards.
+
+## Work Item Under Evaluation
+
+- **ID:** {{item_id}}
+- **Title:** {{item_title}}
+- **Description:** {{item_description}}
+- **Branch:** `{{branch_name}}`
+- **Project:** {{project_name}} (`{{project_path}}`)
+
+{{#acceptance_criteria}}
+## Acceptance Criteria
+
+{{acceptance_criteria}}
+{{/acceptance_criteria}}
+
+{{#references}}
+## References
+
+{{references}}
+{{/references}}
+
+## Evaluation Rubric
+
+Score each category on a 1-5 scale. A category **passes** at 3 or above.
+
+### 1. Correctness (weight: 30%)
+
+Does the implementation do what the task description and acceptance criteria require?
+
+- **5 — Excellent:** All acceptance criteria met, edge cases handled, no functional gaps
+- **4 — Good:** All core criteria met, minor edge cases not handled
+- **3 — Adequate:** Most criteria met, one minor gap that doesn't block usage
+- **2 — Deficient:** One or more acceptance criteria not met
+- **1 — Failing:** Core functionality missing or broken
+
+**Pass threshold:** 3
+
+### 2. Completeness (weight: 25%)
+
+Is the implementation finished end-to-end? No TODO stubs, no half-wired features, no missing integration points.
+
+- **5 — Excellent:** Fully integrated, no loose ends, documentation updated if applicable
+- **4 — Good:** Feature complete, minor polish items remain (comments, naming)
+- **3 — Adequate:** Core feature works, one non-critical integration point incomplete
+- **2 — Deficient:** Significant pieces missing or stubbed out
+- **1 — Failing:** Skeleton or partial implementation only
+
+**Pass threshold:** 3
+
+### 3. Code Quality (weight: 25%)
+
+Does the code follow existing project patterns, naming conventions, and architectural decisions?
+
+- **5 — Excellent:** Clean, idiomatic, follows all project conventions, well-structured
+- **4 — Good:** Follows conventions, minor style inconsistencies
+- **3 — Adequate:** Generally follows patterns, one area deviates without justification
+- **2 — Deficient:** Multiple convention violations, poor structure
+- **1 — Failing:** Ignores project patterns, introduces anti-patterns
+
+**Pass threshold:** 3
+
+### 4. Test Coverage (weight: 20%)
+
+Are there tests for the new functionality? Do existing tests still pass?
+
+- **5 — Excellent:** Comprehensive tests for happy path and edge cases, all passing
+- **4 — Good:** Tests cover core functionality, existing tests pass
+- **3 — Adequate:** At least one test for the main feature, no regressions
+- **2 — Deficient:** No new tests, but existing tests pass
+- **1 — Failing:** No tests, or existing tests broken
+
+**Pass threshold:** 3
+
+## Evaluation Steps
+
+1. **Fetch and review the diff:**
+   ```bash
+   git fetch origin
+   git diff {{main_branch}}...origin/{{branch_name}}
+   ```
+
+2. **Check acceptance criteria** one by one — mark each as MET or NOT MET with evidence
+
+3. **Review code quality** — check for pattern adherence, naming, structure
+
+4. **Verify tests:**
+   ```bash
+   cd {{project_path}}
+   npm test
+   ```
+
+5. **Calculate scores** using the rubric above
+
+6. **Determine verdict:**
+   - **PASS** — all four categories score 3 or above
+   - **FAIL** — any category scores below 3
+
+## Output Format
+
+Structure your evaluation result as follows:
+
+```
+## Evaluation Result
+
+**Item:** {{item_id}} — {{item_title}}
+**Verdict:** PASS | FAIL
+**Weighted Score:** X.X / 5.0
+
+### Scores
+
+| Category | Score | Pass | Notes |
+|----------|-------|------|-------|
+| Correctness | X/5 | YES/NO | ... |
+| Completeness | X/5 | YES/NO | ... |
+| Code Quality | X/5 | YES/NO | ... |
+| Test Coverage | X/5 | YES/NO | ... |
+
+### Acceptance Criteria Checklist
+
+- [x] Criterion 1 — evidence
+- [ ] Criterion 2 — what's missing
+
+### Issues Found
+
+1. **[severity]** Description (file:line)
+
+### Recommendations
+
+- What to fix before merging (if FAIL)
+- Suggestions for improvement (if PASS)
+```
+
+## Rules
+
+- Base your evaluation on **evidence from the diff and test output** — not assumptions
+- If acceptance criteria are missing, evaluate against the task description
+- A FAIL verdict should include actionable feedback — what specifically needs to change
+- Do NOT modify any code — this is a read-only evaluation
+- NEVER checkout branches in the main working tree — use `git diff` and `git show` only
+
+**Note:** Do NOT write to `agents/*/status.json` — the engine manages your status automatically.

package/routing.md

@@ -17,6 +17,7 @@ How the engine decides who handles what. Parsed by engine.js — keep the table
 | test | dallas | ralph |
 | ask | ripley | rebecca |
 | verify | dallas | ralph |
+| evaluate | ripley | rebecca |
 | decompose | ripley | rebecca |
 | meeting | ripley | rebecca |