@sienklogic/plan-build-run 2.22.2 → 2.23.0

package/CHANGELOG.md

@@ -5,6 +5,33 @@ All notable changes to Plan-Build-Run will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [2.23.0](https://github.com/SienkLogic/plan-build-run/compare/plan-build-run-v2.22.2...plan-build-run-v2.23.0) (2026-02-24)
+
+
+### Features
+
+* **28-01:** add local LLM foundation — client, health, metrics, config schema, hook integrations, tests ([44b5a77](https://github.com/SienkLogic/plan-build-run/commit/44b5a773fbb22d77ac0dbf3325201980dd1e9635))
+* **29-01:** integrate local LLM into hooks — artifact classification, task validation, error classification, CLI ([785a708](https://github.com/SienkLogic/plan-build-run/commit/785a708039b0c363589e7151c40df989bb2a7959))
+* **30-01:** add metrics display — session summaries, status skill, CLI, dashboard analytics ([d4ae4a4](https://github.com/SienkLogic/plan-build-run/commit/d4ae4a400932a324f61872169c1dae73cc2923ce))
+* **30-03:** add local-llm-metrics.service.js with getLlmMetrics and Vitest tests ([fc1dd8f](https://github.com/SienkLogic/plan-build-run/commit/fc1dd8f6fbada707543dea848e7355ff52896667))
+* **30-03:** wire getLlmMetrics into /analytics route and add Local LLM Offload section to EJS template ([5fcb61e](https://github.com/SienkLogic/plan-build-run/commit/5fcb61e25848395a4d59f4c555bbb8cf913efb72))
+* **31-01:** add adaptive router — complexity heuristic, confidence gate, 3 routing strategies ([7905462](https://github.com/SienkLogic/plan-build-run/commit/7905462a5dc11063813c1c31400a17f4d2314a23))
+* **32-01:** add agent support — source scoring, error classification, context summarization, prompt injection ([3693660](https://github.com/SienkLogic/plan-build-run/commit/36936609200596fe04cbf27f6eff41566733a74e))
+* **33-01:** add shadow mode, threshold tuner, comprehensive tests, docs, cross-plugin sync ([dbacfed](https://github.com/SienkLogic/plan-build-run/commit/dbacfed98bd6819babad2f07df588cf545d5b99c))
+* **34-01:** add config.features.source_scoring feature flag guard to score-source.js ([a0945a2](https://github.com/SienkLogic/plan-build-run/commit/a0945a2ffd1067ab3644063a8deee4e06a42ea9e))
+* **34-01:** wire runShadow() into router.js post-call path for all 3 routing strategies ([f233948](https://github.com/SienkLogic/plan-build-run/commit/f23394810ab0f647f8eb1ba1d8f78a5eb7d048d8))
+* **35-01:** GREEN - add escapeHtml helper and use it in HTMX error handler path ([a1830c9](https://github.com/SienkLogic/plan-build-run/commit/a1830c9230ba7ab1ea693a7ff2d36e786d5878d3))
+* **35-01:** GREEN - add sanitize-html post-processing to planning.repository ([6d8122b](https://github.com/SienkLogic/plan-build-run/commit/6d8122b83cf1ed62101a3e9fc2f436caf6b0858e))
+* **35-03:** add Quick Tasks view with /quick and /quick/:id routes ([22abe29](https://github.com/SienkLogic/plan-build-run/commit/22abe29e3951ea8e4fbd045116f6216326ee3907))
+
+
+### Documentation
+
+* **34-01:** mark all LLM-01 through LLM-34 traceability entries as Verified ([4311497](https://github.com/SienkLogic/plan-build-run/commit/4311497e1e91f55a575da4468924d21903457cdf))
+* **quick-002:** add .active-skill stale detection to health Check 10 ([3f95b16](https://github.com/SienkLogic/plan-build-run/commit/3f95b16461478bf4271ef4f7ce94582188869bae))
+* **quick-002:** fix NEXT UP banner indentation in milestone SKILL.md ([2983000](https://github.com/SienkLogic/plan-build-run/commit/2983000bd4df5e95c63cdc8f79601c7894883dbb))
+* **quick-002:** replace arrow-list with bullet style in help SKILL.md ([ad78663](https://github.com/SienkLogic/plan-build-run/commit/ad7866375c6f7fd7dda3e195413df95f3ee191bc))
+
 ## [2.22.2](https://github.com/SienkLogic/plan-build-run/compare/plan-build-run-v2.22.1...plan-build-run-v2.22.2) (2026-02-24)
 
 

package/dashboard/package.json

@@ -26,7 +26,8 @@
     "express": "^5.1.0",
     "gray-matter": "^4.0.3",
     "helmet": "^8.1.0",
-    "marked": "^17.0.1"
+    "marked": "^17.0.1",
+    "sanitize-html": "^2.13.0"
   },
   "devDependencies": {
     "@vitest/coverage-v8": "^4.0.18",

package/dashboard/src/middleware/errorHandler.js

@@ -3,6 +3,16 @@
  * MUST have exactly 4 parameters for Express to recognize it as an error handler.
  */
 
+function escapeHtml(str) {
+  if (!str) return '';
+  return str
+    .replace(/&/g, '&')
+    .replace(/</g, '&lt;')
+    .replace(/>/g, '&gt;')
+    .replace(/"/g, '&quot;')
+    .replace(/'/g, '&#39;');
+}
+
 // eslint-disable-next-line no-unused-vars
 export default function errorHandler(err, req, res, next) {
   // If headers already sent, delegate to Express default handler
@@ -40,9 +50,9 @@ export default function errorHandler(err, req, res, next) {
 
   // Render response
   if (isHtmx) {
-    let html = `<h1>Error ${status}</h1><p>${templateData.message}</p>`;
+    let html = `<h1>Error ${status}</h1><p>${escapeHtml(templateData.message)}</p>`;
     if (templateData.stack) {
-      html += `<pre><code>${templateData.stack}</code></pre>`;
+      html += `<pre><code>${escapeHtml(templateData.stack)}</code></pre>`;
     }
     html += '<p><a href="/">Return to Dashboard</a></p>';
     return res.status(status).send(html);

package/dashboard/src/repositories/planning.repository.js

@@ -2,6 +2,7 @@ import { readFile, readdir } from 'node:fs/promises';
 import { join, resolve, relative, normalize } from 'node:path';
 import matter from 'gray-matter';
 import { marked } from 'marked';
+import sanitizeHtml from 'sanitize-html';
 
 marked.setOptions({ gfm: true, breaks: false });
 
@@ -76,9 +77,30 @@ export async function readMarkdownFile(filePath) {
 
   const html = marked.parse(content);
 
+  const sanitizedHtml = sanitizeHtml(html, {
+    allowedTags: sanitizeHtml.defaults.allowedTags.concat([
+      'h1', 'h2', 'h3', 'h4', 'h5', 'h6',
+      'img', 'details', 'summary', 'del', 'ins',
+      'sup', 'sub', 'abbr', 'kbd', 'mark'
+    ]),
+    allowedAttributes: {
+      ...sanitizeHtml.defaults.allowedAttributes,
+      'a': ['href', 'name', 'target', 'rel'],
+      'img': ['src', 'alt', 'title', 'width', 'height'],
+      'code': ['class'],
+      'pre': ['class'],
+      'span': ['class'],
+      'div': ['class'],
+      'td': ['align'],
+      'th': ['align', 'scope'],
+      'h1': ['id'], 'h2': ['id'], 'h3': ['id'],
+      'h4': ['id'], 'h5': ['id'], 'h6': ['id']
+    }
+  });
+
   return {
     frontmatter: data,
-    html,
+    html: sanitizedHtml,
     rawContent: content
   };
 }

package/dashboard/src/routes/pages.routes.js

@@ -5,7 +5,9 @@ import { parseStateFile, derivePhaseStatuses } from '../services/dashboard.servi
 import { listPendingTodos, getTodoDetail, createTodo, completeTodo } from '../services/todo.service.js';
 import { getAllMilestones, getMilestoneDetail } from '../services/milestone.service.js';
 import { getProjectAnalytics } from '../services/analytics.service.js';
+import { getLlmMetrics } from '../services/local-llm-metrics.service.js';
 import { listNotes } from '../services/notes.service.js';
+import { listQuickTasks, getQuickTask } from '../services/quick.service.js';
 
 const router = Router();
 
@@ -358,14 +360,18 @@ router.get('/dependencies', async (req, res) => {
 
 router.get('/analytics', async (req, res) => {
   const projectDir = req.app.locals.projectDir;
-  const analytics = await getProjectAnalytics(projectDir);
+  const [analytics, llmMetrics] = await Promise.all([
+    getProjectAnalytics(projectDir),
+    getLlmMetrics(projectDir)
+  ]);
 
   const templateData = {
     title: 'Analytics',
     activePage: 'analytics',
     currentPath: '/analytics',
     breadcrumbs: [{ label: 'Analytics' }],
-    analytics
+    analytics,
+    llmMetrics
   };
 
   res.setHeader('Vary', 'HX-Request');
@@ -423,4 +429,61 @@ router.get('/roadmap', async (req, res) => {
   }
 });
 
+router.get('/quick', async (req, res) => {
+  const projectDir = req.app.locals.projectDir;
+  const tasks = await listQuickTasks(projectDir);
+
+  const templateData = {
+    title: 'Quick Tasks',
+    activePage: 'quick',
+    currentPath: '/quick',
+    breadcrumbs: [{ label: 'Quick Tasks' }],
+    tasks
+  };
+
+  res.setHeader('Vary', 'HX-Request');
+
+  if (req.get('HX-Request') === 'true') {
+    res.render('partials/quick-content', templateData);
+  } else {
+    res.render('quick', templateData);
+  }
+});
+
+router.get('/quick/:id', async (req, res) => {
+  const { id } = req.params;
+
+  // Validate ID format: must be exactly three digits
+  if (!/^\d{3}$/.test(id)) {
+    const err = new Error('Quick Task ID must be a three-digit number (e.g., 001, 005, 042)');
+    err.status = 404;
+    throw err;
+  }
+
+  const projectDir = req.app.locals.projectDir;
+  const task = await getQuickTask(projectDir, id);
+
+  if (!task) {
+    const err = new Error(`Quick task ${id} not found`);
+    err.status = 404;
+    throw err;
+  }
+
+  const templateData = {
+    title: `Quick Task ${task.id}: ${task.title}`,
+    activePage: 'quick',
+    currentPath: '/quick/' + id,
+    breadcrumbs: [{ label: 'Quick Tasks', url: '/quick' }, { label: 'Task ' + id }],
+    ...task
+  };
+
+  res.setHeader('Vary', 'HX-Request');
+
+  if (req.get('HX-Request') === 'true') {
+    res.render('partials/quick-detail-content', templateData);
+  } else {
+    res.render('quick-detail', templateData);
+  }
+});
+
 export default router;

package/dashboard/src/services/local-llm-metrics.service.js

@@ -0,0 +1,81 @@
+import { readFile } from 'node:fs/promises';
+import { join } from 'node:path';
+
+/**
+ * Read and aggregate local LLM metrics from .planning/logs/local-llm-metrics.jsonl.
+ *
+ * @param {string} projectDir - Absolute path to the project root
+ * @returns {Promise<{summary: object, byOperation: Array, baseline: object}|null>}
+ *   Returns null if the file does not exist, is empty, or has no valid entries.
+ */
+export async function getLlmMetrics(projectDir) {
+  try {
+    const filePath = join(projectDir, '.planning', 'logs', 'local-llm-metrics.jsonl');
+
+    let raw;
+    try {
+      raw = await readFile(filePath, 'utf8');
+    } catch (err) {
+      if (err.code === 'ENOENT') return null;
+      throw err;
+    }
+
+    // Parse valid JSON lines, skip malformed
+    const entries = [];
+    for (const line of raw.split('\n')) {
+      const trimmed = line.trim();
+      if (!trimmed) continue;
+      try {
+        entries.push(JSON.parse(trimmed));
+      } catch {
+        // skip malformed lines
+      }
+    }
+
+    if (entries.length === 0) return null;
+
+    // Compute aggregate summary
+    const total_calls = entries.length;
+    const fallback_count = entries.filter(e => e.fallback_used === true).length;
+    const fallback_rate_pct = Math.round(fallback_count / total_calls * 100);
+    const avg_latency_ms = Math.round(
+      entries.reduce((sum, e) => sum + (e.latency_ms || 0), 0) / total_calls
+    );
+    const tokens_saved = entries.reduce((sum, e) => sum + (e.tokens_saved_frontier || 0), 0);
+    const cost_saved_usd = parseFloat((tokens_saved * (3.0 / 1_000_000)).toFixed(4));
+
+    const summary = {
+      total_calls,
+      fallback_count,
+      fallback_rate_pct,
+      avg_latency_ms,
+      tokens_saved,
+      cost_saved_usd
+    };
+
+    // Compute byOperation: group entries by operation field
+    const opMap = new Map();
+    for (const e of entries) {
+      const op = e.operation || 'unknown';
+      if (!opMap.has(op)) {
+        opMap.set(op, { operation: op, calls: 0, fallbacks: 0, tokens_saved: 0 });
+      }
+      const rec = opMap.get(op);
+      rec.calls += 1;
+      if (e.fallback_used === true) rec.fallbacks += 1;
+      rec.tokens_saved += e.tokens_saved_frontier || 0;
+    }
+
+    const byOperation = Array.from(opMap.values()).sort((a, b) => b.calls - a.calls);
+
+    // Compute baseline (LLM-15)
+    const baseline = {
+      hook_invocations: total_calls,
+      estimated_frontier_tokens_without_local: tokens_saved
+    };
+
+    return { summary, byOperation, baseline };
+  } catch {
+    return null;
+  }
+}

package/dashboard/src/services/quick.service.js

@@ -0,0 +1,62 @@
+import { readdir } from 'node:fs/promises';
+import { join } from 'node:path';
+import { readMarkdownFile } from '../repositories/planning.repository.js';
+
+export async function listQuickTasks(projectDir) {
+  const quickDir = join(projectDir, '.planning', 'quick');
+  let entries;
+  try {
+    entries = await readdir(quickDir, { withFileTypes: true });
+  } catch (err) {
+    if (err.code === 'ENOENT') return [];
+    throw err;
+  }
+
+  const dirs = entries.filter(e => e.isDirectory()).map(e => e.name).sort();
+  const tasks = [];
+  for (const dirName of dirs) {
+    const match = dirName.match(/^(\d{3})-(.+)$/);
+    if (!match) continue;
+    const [, id, slug] = match;
+    const title = slug.split('-').map(w => w.charAt(0).toUpperCase() + w.slice(1)).join(' ');
+    let status = 'in-progress';
+    try {
+      const summaryPath = join(quickDir, dirName, 'SUMMARY.md');
+      const { frontmatter } = await readMarkdownFile(summaryPath);
+      if (frontmatter.status) status = frontmatter.status;
+    } catch { /* No SUMMARY.md yet */ }
+    tasks.push({ id, slug, dirName, title, status });
+  }
+  return tasks;
+}
+
+export async function getQuickTask(projectDir, id) {
+  const quickDir = join(projectDir, '.planning', 'quick');
+  let entries;
+  try {
+    entries = await readdir(quickDir, { withFileTypes: true });
+  } catch (err) {
+    if (err.code === 'ENOENT') return null;
+    throw err;
+  }
+  const dirEntry = entries.filter(e => e.isDirectory()).find(e => e.name.startsWith(id + '-'));
+  if (!dirEntry) return null;
+
+  const taskDir = join(quickDir, dirEntry.name);
+  const match = dirEntry.name.match(/^(\d{3})-(.+)$/);
+  const slug = match ? match[2] : dirEntry.name;
+  const title = slug.split('-').map(w => w.charAt(0).toUpperCase() + w.slice(1)).join(' ');
+
+  let planHtml = null, summaryHtml = null, summaryFrontmatter = {};
+  try {
+    const plan = await readMarkdownFile(join(taskDir, 'PLAN.md'));
+    planHtml = plan.html;
+  } catch { /* missing */ }
+  try {
+    const summary = await readMarkdownFile(join(taskDir, 'SUMMARY.md'));
+    summaryHtml = summary.html;
+    summaryFrontmatter = summary.frontmatter;
+  } catch { /* missing */ }
+
+  return { id, slug, dirName: dirEntry.name, title, status: summaryFrontmatter.status || 'in-progress', planHtml, summaryHtml };
+}

package/dashboard/src/views/partials/analytics-content.ejs

@@ -88,3 +88,64 @@
 <% } else { %>
 <p>No phase data available.</p>
 <% } %>
+
+<% if (typeof llmMetrics !== 'undefined' && llmMetrics) { %>
+<article>
+  <header>Local LLM Offload</header>
+  <div class="grid">
+    <article>
+      <header>Total Calls</header>
+      <strong class="stat-value"><%= llmMetrics.summary.total_calls %></strong>
+      <span class="stat-unit">calls</span>
+    </article>
+    <article>
+      <header>Tokens Saved</header>
+      <strong class="stat-value"><%= llmMetrics.summary.tokens_saved.toLocaleString() %></strong>
+      <span class="stat-unit">frontier tokens</span>
+    </article>
+    <article>
+      <header>Est. Cost Saved</header>
+      <strong class="stat-value">$<%= llmMetrics.summary.cost_saved_usd.toFixed(4) %></strong>
+      <span class="stat-unit">at $3/M tokens</span>
+    </article>
+    <article>
+      <header>Fallback Rate</header>
+      <strong class="stat-value"><%= llmMetrics.summary.fallback_rate_pct %>%</strong>
+      <span class="stat-unit"><%= llmMetrics.summary.fallback_count %> fallbacks</span>
+    </article>
+    <article>
+      <header>Avg Latency</header>
+      <strong class="stat-value"><%= llmMetrics.summary.avg_latency_ms %></strong>
+      <span class="stat-unit">ms/call</span>
+    </article>
+  </div>
+  <% if (llmMetrics.byOperation && llmMetrics.byOperation.length > 0) { %>
+  <div class="overflow-auto" style="margin-top: var(--space-md);">
+    <table>
+      <thead>
+        <tr>
+          <th>Operation</th>
+          <th>Calls</th>
+          <th>Fallbacks</th>
+          <th>Tokens Saved</th>
+        </tr>
+      </thead>
+      <tbody>
+        <% llmMetrics.byOperation.forEach(op => { %>
+        <tr>
+          <td><%= op.operation %></td>
+          <td><%= op.calls %></td>
+          <td><%= op.fallbacks %></td>
+          <td><%= op.tokens_saved.toLocaleString() %></td>
+        </tr>
+        <% }) %>
+      </tbody>
+    </table>
+  </div>
+  <% } %>
+  <footer style="color: var(--pico-muted-color); font-size: 0.85em;">
+    Baseline estimate: each local call replaced ~<%= llmMetrics.baseline.estimated_frontier_tokens_without_local.toLocaleString() %> frontier tokens total.
+    Advisory only — no data collected when local LLM is disabled.
+  </footer>
+</article>
+<% } %>

package/dashboard/src/views/partials/quick-content.ejs

@@ -0,0 +1,40 @@
+<%- include('breadcrumbs', { breadcrumbs: typeof breadcrumbs !== 'undefined' ? breadcrumbs : [] }) %>
+<h1>Quick Tasks</h1>
+
+<% if (typeof tasks !== 'undefined' && tasks.length > 0) { %>
+<article>
+  <div class="table-wrap">
+  <table>
+    <thead>
+      <tr>
+        <th scope="col">ID</th>
+        <th scope="col">Title</th>
+        <th scope="col">Status</th>
+      </tr>
+    </thead>
+    <tbody>
+      <% tasks.forEach(function(task) { %>
+      <tr>
+        <td><%= task.id %></td>
+        <td>
+          <a href="/quick/<%= task.id %>"
+             hx-get="/quick/<%= task.id %>"
+             hx-target="#main-content"
+             hx-push-url="true">
+            <%= task.title %>
+          </a>
+        </td>
+        <td>
+          <span class="status-badge" data-status="<%= task.status %>">
+            <%= task.status %>
+          </span>
+        </td>
+      </tr>
+      <% }); %>
+    </tbody>
+  </table>
+  </div>
+</article>
+<% } else { %>
+<%- include('empty-state', { icon: '⚡', title: 'No quick tasks found', action: '' }) %>
+<% } %>

package/dashboard/src/views/partials/quick-detail-content.ejs

@@ -0,0 +1,29 @@
+<%- include('breadcrumbs', { breadcrumbs: typeof breadcrumbs !== 'undefined' ? breadcrumbs : [] }) %>
+<h1><%= title %></h1>
+
+<p><a href="/quick">&larr; Back to Quick Tasks</a></p>
+
+<article>
+  <header>
+    <strong>Quick Task <%= id %></strong>
+    &nbsp;
+    <span class="status-badge" data-status="<%= status %>">
+      <%= status %>
+    </span>
+  </header>
+
+  <% if (planHtml) { %>
+  <section>
+    <h2>Plan</h2>
+    <%- planHtml %>
+  </section>
+  <% } %>
+
+  <% if (summaryHtml) { %>
+  <hr>
+  <section>
+    <h2>Summary</h2>
+    <%- summaryHtml %>
+  </section>
+  <% } %>
+</article>

package/dashboard/src/views/partials/sidebar.ejs

@@ -79,6 +79,14 @@
             Notes
           </a>
         </li>
+        <li>
+          <a href="/quick"
+             hx-get="/quick"
+             hx-target="#main-content"
+             hx-push-url="true"<%= typeof activePage !== 'undefined' && activePage === 'quick' ? ' aria-current="page"' : '' %>>
+            Quick Tasks
+          </a>
+        </li>
       </ul>
     </details>
 

package/dashboard/src/views/quick-detail.ejs

@@ -0,0 +1,5 @@
+<%- include('partials/layout-top', { title: title, activePage: 'quick' }) %>
+
+<%- include('partials/quick-detail-content') %>
+
+<%- include('partials/layout-bottom') %>

package/dashboard/src/views/quick.ejs

@@ -0,0 +1,5 @@
+<%- include('partials/layout-top', { title: 'Quick Tasks', activePage: 'quick' }) %>
+
+<%- include('partials/quick-content') %>
+
+<%- include('partials/layout-bottom') %>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sienklogic/plan-build-run",
-  "version": "2.22.2",
+  "version": "2.23.0",
   "description": "Plan it, Build it, Run it — structured development workflow for Claude Code",
   "keywords": [
     "claude-code",

package/plugins/copilot-pbr/agents/debugger.agent.md

@@ -138,6 +138,21 @@ Then emit a `DECISION` checkpoint asking the user to approve, modify, or reject
 
 **Commit format**: `fix({scope}): {description}` with body: `Root cause: ...` and `Debug session: .planning/debug/{slug}.md`
 
+## Local LLM Error Classification (Optional)
+
+When you receive an error message or stack trace, you MAY use the local LLM to classify it before starting hypothesis generation. This is advisory — skip it if unavailable.
+
+```bash
+# Write the error to a temp file, then classify:
+echo "Error text here" > /tmp/debug-error.txt
+node "${PLUGIN_ROOT}/scripts/pbr-tools.js" llm classify-error /tmp/debug-error.txt debugger 2>/dev/null
+# Returns: {"category":"missing_output","confidence":0.91,"latency_ms":1840,"fallback_used":false}
+```
+
+Categories: `connection_refused`, `timeout`, `missing_output`, `wrong_output_format`, `permission_error`, `unknown`.
+
+If classification succeeds, use the returned category to bias your initial hypothesis ranking. If it returns null or fails, proceed with manual hypothesis generation as normal.
+
 ## Common Bug Patterns
 
 Reference: `references/common-bug-patterns.md` — covers off-by-one, null/undefined, async/timing, state management, import/module, environment, and data shape patterns.

package/plugins/copilot-pbr/agents/researcher.agent.md

@@ -54,6 +54,26 @@ All claims must be attributed to a source level. Higher levels override lower le
 
 **Offline Fallback**: If web tools are unavailable (air-gapped environment, MCP not configured), rely on local sources: codebase analysis via Glob/Grep, existing documentation, and README files. Assign these S3-S4 confidence levels. Do not attempt WebFetch or WebSearch — note in the output header that external sources were unavailable.
 
+## Local LLM Source Scoring (Optional)
+
+If local LLM offload is configured, you MAY use it to score source credibility instead of manually assigning S-levels. This is advisory — never wait on it or fail if it returns null.
+
+Check availability first:
+
+```bash
+node "${PLUGIN_ROOT}/scripts/pbr-tools.js" llm status 2>/dev/null
+```
+
+If `enabled: true`, score a source excerpt:
+
+```bash
+echo "Source URL and content excerpt" > /tmp/source-excerpt.txt
+node "${PLUGIN_ROOT}/scripts/pbr-tools.js" llm score-source "https://example.com/docs" /tmp/source-excerpt.txt 2>/dev/null
+# Returns: {"level":"S2","confidence":0.87,"reason":"Official library documentation page"}
+```
+
+Use the returned `level` to set your source tag. If the call fails or returns `null`, assign the level manually per the hierarchy table above.
+
 ---
 
 ## Confidence Levels

package/plugins/copilot-pbr/agents/synthesizer.agent.md

@@ -98,6 +98,18 @@ conflicts: N
 - **Research gaps**: Add `[RESEARCH GAP]` flag, add to Open Questions with high impact, never fabricate
 - **Duplicates**: Consolidate into one entry, note multi-source agreement, reference all documents
 
+## Local LLM Context Summarization (Optional)
+
+When input research documents are large (>2000 words combined), you MAY use the local LLM to pre-summarize each document before synthesis. This reduces your own context consumption. Advisory only — if unavailable, read documents normally.
+
+```bash
+# Pre-summarize a large research document to ~150 words:
+node "${PLUGIN_ROOT}/scripts/pbr-tools.js" llm summarize /path/to/RESEARCH.md 150 2>/dev/null
+# Returns: {"summary":"...plain text summary under 150 words...","latency_ms":2100,"fallback_used":false}
+```
+
+Use the returned `summary` string as your working copy of that document's findings. Still read the original for any specific version numbers, code examples, or direct quotes needed in the output.
+
 ## Anti-Patterns
 
 ### Universal Anti-Patterns

package/plugins/copilot-pbr/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "pbr",
   "displayName": "Plan-Build-Run",
-  "version": "2.22.2",
+  "version": "2.23.0",
   "description": "Plan-Build-Run — Structured development workflow for GitHub Copilot CLI. Solves context rot through disciplined agent delegation, structured planning, atomic execution, and goal-backward verification.",
   "author": {
     "name": "SienkLogic",