@sienklogic/plan-build-run 2.19.1 → 2.19.2

package/CHANGELOG.md

@@ -2,6 +2,45 @@
 
 All notable changes to Plan-Build-Run will be documented in this file.
 
+## [2.19.2](https://github.com/SienkLogic/plan-build-run/compare/plan-build-run-v2.19.1...plan-build-run-v2.19.2) — 2026-03-24
+
+### CI/CD
+
+* Remove double blank line in config-reference.md (markdownlint MD012) ([f72d4f2f](https://github.com/SienkLogic/plan-build-run/commit/f72d4f2f))
+* Wire spot-check CLI command and fix empty block lint errors ([b4abfc2c](https://github.com/SienkLogic/plan-build-run/commit/b4abfc2c))
+
+### CLI Tools
+
+* Repair pbr-tools.test.js brace nesting from Phase 141 legacy removal ([031c0d06](https://github.com/SienkLogic/plan-build-run/commit/031c0d06))
+
+### Configuration
+
+* Add missing feature flags to schema and fix default drift ([248e5b5a](https://github.com/SienkLogic/plan-build-run/commit/248e5b5a))
+
+### Dashboard
+
+* Fix log filename, health require paths, version, and research date ([b0eadc4a](https://github.com/SienkLogic/plan-build-run/commit/b0eadc4a))
+
+### Hooks
+
+* Suppress git diff --cached stderr output with stdio pipe ([7cec470b](https://github.com/SienkLogic/plan-build-run/commit/7cec470b))
+* Restrict dashboard CORS to localhost and add hook event log rotation ([57e829dd](https://github.com/SienkLogic/plan-build-run/commit/57e829dd))
+* Remove leftover enrichCommitLlm reference from pre-bash-dispatch ([ffa1ba03](https://github.com/SienkLogic/plan-build-run/commit/ffa1ba03))
+* Add PreToolUse response translation layer in hook-server.js ([b8cbbef6](https://github.com/SienkLogic/plan-build-run/commit/b8cbbef6))
+
+### Skills
+
+* Replace stale plan-build-run/bin/lib/ paths and fix state-update line limit ([a67d6c4b](https://github.com/SienkLogic/plan-build-run/commit/a67d6c4b))
+
+### Testing
+
+* Update suggest-compact test for legacy .compact-counter removal ([f9e35aeb](https://github.com/SienkLogic/plan-build-run/commit/f9e35aeb))
+
+### Other
+
+* Remove orphaned config keys and wrap git diff --cached in try-catch ([4d863ecd](https://github.com/SienkLogic/plan-build-run/commit/4d863ecd))
+* Add PBR_DEBUG logging to empty catches and fix 9 .cjs extension mismatches in verify.js ([1931893d](https://github.com/SienkLogic/plan-build-run/commit/1931893d))
+
 ## [2.19.1](https://github.com/SienkLogic/plan-build-run/compare/plan-build-run-v2.19.0...plan-build-run-v2.19.1) — 2026-03-23
 
 ### Skills

package/CLAUDE.md

@@ -50,17 +50,17 @@ Three layers:
 
 Markdown files with YAML frontmatter defining slash commands (`/pbr:new-project`, `/pbr:plan-phase`, etc.). Each SKILL.md is a complete prompt that tells the orchestrator what to do. Skills read state, interact with the user, and spawn agents.
 
-34 skills: audit, begin, build, config, continue, dashboard, debug, discuss, do, explore, health, help, import, intel, milestone, note, pause, plan, profile, quick, resume, review, scan, setup, status, statusline, test, todo, undo, and more.
+46 skills: audit, begin, build, config, continue, dashboard, debug, discuss, do, explore, health, help, import, intel, milestone, note, pause, plan, profile, quick, resume, review, scan, setup, status, statusline, test, todo, undo, and more.
 
 ### Agents (`agents/{name}.md`)
 
 Markdown files with YAML frontmatter defining specialized subagent prompts. Agents run in fresh `Task()` contexts with clean 200k token windows. Spawned via `subagent_type: "pbr:{name}"` — auto-loaded by Claude Code.
 
-17 agents: audit, codebase-mapper, debugger, dev-sync, executor, general, integration-checker, intel-updater, nyquist-auditor, plan-checker, planner, researcher, roadmapper, synthesizer, ui-checker, ui-researcher, verifier.
+18 agents: audit, codebase-mapper, debugger, dev-sync, executor, general, integration-checker, intel-updater, nyquist-auditor, plan-checker, planner, researcher, roadmapper, synthesizer, ui-checker, ui-researcher, verifier.
 
-### Hook Scripts (`hooks/*.js`)
+### Hook Scripts (`plugins/pbr/scripts/*.js`)
 
-50 Node.js hook scripts that fire on Claude Code lifecycle events. Configured in `hooks/hooks.json`. All use CommonJS, must be cross-platform (`path.join()`, not hardcoded separators), and log via `logHook()` from `hook-logger.js`.
+66 Node.js hook scripts that fire on Claude Code lifecycle events. Configured in `plugins/pbr/hooks/hooks.json`. All use CommonJS, must be cross-platform (`path.join()`, not hardcoded separators), and log via `logHook()` from `hook-logger.js`.
 
 **Dispatch pattern**: Several hooks use dispatch scripts that fan out to sub-scripts based on the file being written/read:
 
@@ -68,22 +68,33 @@ Markdown files with YAML frontmatter defining specialized subagent prompts. Agen
 |------------|-------------|-------------|
 | SessionStart | progress-tracker.js | — (injects project state) |
 | PostToolUse (Write\|Edit) | post-write-dispatch.js | check-plan-format.js, check-roadmap-sync.js, check-state-sync.js |
-| PostToolUse (Write\|Edit) | post-write-quality.js | — (autoFormat, typeCheck, detectConsoleLogs) |
-| PostToolUse (Task) | check-subagent-output.js | — (validates agent output) |
+| PostToolUse (Write\|Edit) | context-bridge.js | — (context tier tracking) |
+| PostToolUse (Write\|Edit) | graph-update.js | — (architecture graph) |
+| PostToolUse (Write\|Edit) | architecture-guard.js | — (dependency violations) |
 | PostToolUse (Write\|Edit) | suggest-compact.js | — (context budget warnings) |
-| PostToolUse (Read) | track-context-budget.js | — (tracks reads for budget) |
+| PostToolUse (Read\|Glob\|Grep) | track-context-budget.js | — (tracks reads for budget) |
+| PostToolUse (Bash) | post-bash-triage.js | — (test output triage) |
+| PostToolUse (Bash) | context-bridge.js | — (context tier tracking) |
+| PostToolUse (Task) | check-subagent-output.js | — (validates agent output) |
+| PostToolUse (Task) | context-bridge.js | — (context tier tracking) |
+| PostToolUse (AskUserQuestion) | track-user-gates.js | — (gate tracking) |
 | PostToolUseFailure | log-tool-failure.js | — (logs failures) |
 | PreToolUse (Bash) | pre-bash-dispatch.js | validate-commit.js, check-dangerous-commands.js |
 | PreToolUse (Write\|Edit) | pre-write-dispatch.js | check-skill-workflow.js, check-summary-gate.js, check-doc-sprawl.js |
+| PreToolUse (Task) | pre-task-dispatch.js | — (build/plan gates) |
+| PreToolUse (Skill) | pre-skill-dispatch.js | — (context budget, args) |
+| PreToolUse (Read) | block-skill-self-read.js | — (prevents SKILL.md self-read) |
+| PreToolUse (EnterPlanMode) | intercept-plan-mode.js | — (blocks plan mode in PBR) |
 | PreCompact | context-budget-check.js | — (preserves STATE.md) |
+| PostCompact | post-compact.js | — (post-compact recovery) |
 | Stop | auto-continue.js | — (chains next command) |
-| SubagentStart/Stop | log-subagent.js | — (tracks lifecycle) |
-| SubagentStop | event-handler.js | — (auto-verification trigger) |
+| SubagentStart | log-subagent.js | — (tracks lifecycle) |
+| SubagentStop | log-subagent.js, event-handler.js | — (auto-verification trigger) |
 | TaskCompleted | task-completed.js | — (processes task completion) |
-| InstructionsLoaded | hook-server-client.js | — (detects instruction reload) |
-| ConfigChange | hook-server-client.js | — (config change detection) |
-| WorktreeCreate | hook-server-client.js | — (worktree setup) |
-| WorktreeRemove | hook-server-client.js | — (worktree cleanup) |
+| InstructionsLoaded | instructions-loaded.js | — (instruction reload detection) |
+| ConfigChange | check-config-change.js | — (config change detection) |
+| WorktreeCreate | worktree-create.js | — (worktree setup) |
+| WorktreeRemove | worktree-remove.js | — (worktree cleanup) |
 | Notification | log-notification.js | — (notification logging) |
 | UserPromptSubmit | prompt-routing.js | — (prompt routing) |
 | SessionEnd | session-cleanup.js | — (cleanup) |
@@ -99,8 +110,8 @@ Markdown files with YAML frontmatter defining specialized subagent prompts. Agen
 - **`plan-build-run/bin/`** — CLI tools (`pbr-tools.js` + `lib/` modules)
 - **`plan-build-run/references/`** — Shared reference docs loaded by skills (plan format, commit conventions, UI formatting, deviation rules)
 - **`plan-build-run/templates/`** — EJS-style `.tmpl` files for generated markdown (VERIFICATION.md, SUMMARY.md, etc.)
-- **`plugins/pbr/commands/`** — 63 command registration files (one `.md` per command mapping to its skill)
-- **`plan-build-run/skills/shared/`** — 12 shared skill fragments extracted from repeated patterns across skills
+- **`plugins/pbr/commands/`** — 71 command registration files (one `.md` per command mapping to its skill)
+- **`plan-build-run/skills/shared/`** — 15 shared skill fragments extracted from repeated patterns across skills
 - **`plugins/`** — Derivative plugins (codex-pbr, cursor-pbr, copilot-pbr)
 - **`dashboard/`** — Vite + React 18 dashboard with Express backend
 
@@ -125,10 +136,10 @@ argument-hint: "<N> [--flag]"
 ---
 name: agent-name
 description: "What this agent does"
-model: sonnet|inherit|haiku
 memory: none|user|project
 isolation: worktree        # Run agent in isolated git worktree (optional)
 color: <name|hex>          # Terminal color for agent output (optional)
+permissionMode: default    # Optional: permissions for agent (optional)
 tools:
   - Read
   - Write
@@ -136,6 +147,8 @@ tools:
 ---
 ```
 
+Model selection is config-driven via `config.json` `models` map, not via agent frontmatter.
+
 ## Data Flow
 
 Skills and agents communicate through files on disk, not messages:

package/README.md

@@ -145,11 +145,11 @@ Plans are grouped into **waves** based on dependencies. Within each wave, plans
 <details>
 <summary><strong>Three layers: Skills → Agents → Hooks</strong></summary>
 
-### Skills (39 slash commands)
+### Skills (46 slash commands)
 
 Markdown files with YAML frontmatter defining `/pbr:*` slash commands. Each skill is a complete prompt that reads state, interacts with the user, and spawns agents. Skills are the user-facing interface.
 
-### Agents (17 specialized subagents)
+### Agents (18 specialized subagents)
 
 Markdown files defining agent prompts that run in fresh `Task()` contexts with clean 200k token windows. Each agent type has a specific role:
 
@@ -526,7 +526,7 @@ CI runs on Node 18/20/22 across Windows, macOS, and Linux (9 platform combinatio
 
 <div align="center">
 
-**39 skills &bull; 17 agents &bull; 26 hooks &bull; 38 server routes &bull; 4 platforms**
+**46 skills &bull; 18 agents &bull; 26 hooks &bull; 38 server routes &bull; 4 platforms**
 
 **Claude Code is powerful. PBR makes it reliable.**
 

package/dashboard/server/index.js

@@ -44,7 +44,16 @@ function createApp(options = {}) {
   const app = express();
 
   // Core middleware
-  app.use(cors());
+  app.use(cors({
+    origin: (origin, callback) => {
+      // Allow requests with no origin (same-origin, curl, etc.) or localhost
+      if (!origin || /^https?:\/\/(localhost|127\.0\.0\.1)(:\d+)?$/.test(origin)) {
+        callback(null, true);
+      } else {
+        callback(new Error('Not allowed by CORS'));
+      }
+    }
+  }));
   app.use(express.json());
 
   // Store resolved options on app for downstream middleware

package/dashboard/server/routes/agents.js

@@ -5,6 +5,27 @@ const fs = require('fs');
 const path = require('path');
 const { parseFrontmatter } = require('../lib/frontmatter');
 
+/**
+ * Get the path to the current hook log file.
+ * Prefers today's dated file, falls back to most recent, then legacy hooks.jsonl.
+ */
+function getHookLogPath(planningDir) {
+  const today = new Date().toISOString().slice(0, 10);
+  const dated = path.join(planningDir, 'logs', `hooks-${today}.jsonl`);
+  if (fs.existsSync(dated)) return dated;
+  // Fallback: find most recent hooks-*.jsonl
+  const logsDir = path.join(planningDir, 'logs');
+  try {
+    const files = fs.readdirSync(logsDir)
+      .filter(f => /^hooks-\d{4}-\d{2}-\d{2}\.jsonl$/.test(f))
+      .sort()
+      .reverse();
+    if (files.length > 0) return path.join(logsDir, files[0]);
+  } catch (_e) { /* no logs dir */ }
+  // Last fallback: legacy filename
+  return path.join(planningDir, 'logs', 'hooks.jsonl');
+}
+
 /**
  * Agents route factory.
  * Reads agent definitions from plugins/pbr/agents/*.md.
@@ -143,7 +164,7 @@ function createAgentsRouter({ planningDir, agentsDir }) {
    */
   router.get('/hooks', async (_req, res) => {
     try {
-      const eventsPath = path.join(planningDir, 'logs', 'hooks.jsonl');
+      const eventsPath = getHookLogPath(planningDir);
       let events = [];
 
       try {
@@ -179,7 +200,7 @@ function createAgentsRouter({ planningDir, agentsDir }) {
    */
   router.get('/errors', async (_req, res) => {
     try {
-      const eventsPath = path.join(planningDir, 'logs', 'hooks.jsonl');
+      const eventsPath = getHookLogPath(planningDir);
       let errors = [];
 
       try {

package/dashboard/server/routes/health.js

@@ -5,6 +5,9 @@ const path = require('path');
 
 const router = express.Router();
 const startTime = Date.now();
+const pbrVersion = (() => {
+  try { return require('../../../../package.json').version; } catch (_e) { return 'unknown'; }
+})();
 
 /**
  * Resolve the planning directory from request app config.
@@ -35,7 +38,7 @@ router.get('/', (_req, res) => {
   res.json({
     status: 'ok',
     uptime: Math.floor((Date.now() - startTime) / 1000),
-    version: '4.0.0',
+    version: pbrVersion,
     timestamp: new Date().toISOString(),
   });
 });
@@ -53,17 +56,17 @@ router.get('/features', (req, res) => {
     const featureChecks = [
       {
         name: 'progress_visualization',
-        module: '../../../plan-build-run/bin/lib/progress-visualization.cjs',
+        module: '../../../plugins/pbr/scripts/lib/progress-visualization.js',
         fn: 'getProgressData',
       },
       {
         name: 'contextual_help',
-        module: '../../../plan-build-run/bin/lib/contextual-help.cjs',
+        module: '../../../plugins/pbr/scripts/lib/contextual-help.js',
         fn: 'getContextualHelp',
       },
       {
         name: 'team_onboarding',
-        module: '../../../plan-build-run/bin/lib/onboarding-generator.cjs',
+        module: '../../../plugins/pbr/scripts/lib/onboarding-generator.js',
         fn: 'generateOnboardingGuide',
       },
     ];

package/dashboard/server/routes/telemetry.js

@@ -4,6 +4,25 @@ const express = require('express');
 const fs = require('fs');
 const path = require('path');
 
+/**
+ * Get the path to the current hook log file.
+ * Prefers today's dated file, falls back to most recent, then legacy hooks.jsonl.
+ */
+function getHookLogPath(planningDir) {
+  const today = new Date().toISOString().slice(0, 10);
+  const dated = path.join(planningDir, 'logs', `hooks-${today}.jsonl`);
+  if (fs.existsSync(dated)) return dated;
+  const logsDir = path.join(planningDir, 'logs');
+  try {
+    const files = fs.readdirSync(logsDir)
+      .filter(f => /^hooks-\d{4}-\d{2}-\d{2}\.jsonl$/.test(f))
+      .sort()
+      .reverse();
+    if (files.length > 0) return path.join(logsDir, files[0]);
+  } catch (_e) { /* no logs dir */ }
+  return path.join(planningDir, 'logs', 'hooks.jsonl');
+}
+
 /**
  * Telemetry route factory.
  * Reads hook event log and telemetry data from .planning/.
@@ -21,7 +40,7 @@ function createTelemetryRouter({ planningDir }) {
    */
   router.get('/', async (_req, res) => {
     try {
-      const eventsPath = path.join(planningDir, 'logs', 'hooks.jsonl');
+      const eventsPath = getHookLogPath(planningDir);
       let events = [];
 
       try {

package/dashboard/server/services/planning-reader.js

@@ -64,7 +64,7 @@ class PlanningReader {
   /**
    * Read STATE.md and return session state with progress and history.
    * Returns frontmatter fields plus nested progress object and history text.
-   * Falls back to legacy HISTORY.md if STATE.md has no ## History section.
+   * Reads ## History section from STATE.md body.
    */
   async getStatus() {
     const content = await safeReadFile(path.join(this.planningDir, 'STATE.md'));
@@ -76,13 +76,6 @@ class PlanningReader {
     const historyMatch = body.match(/## History\n([\s\S]*?)(?=\n## |\s*$)/);
     if (historyMatch) {
       history = historyMatch[1].trim();
-    } else {
-      // Backwards compat: read legacy HISTORY.md if it exists
-      const legacyHistory = await safeReadFile(path.join(this.planningDir, 'HISTORY.md'));
-      if (legacyHistory) {
-        const { body: histBody } = parseFrontmatter(legacyHistory);
-        history = histBody.trim() || null;
-      }
     }
 
     // Build response with progress as nested object (parsed by frontmatter lib)
@@ -112,7 +105,7 @@ class PlanningReader {
   /**
    * Read PROJECT.md and extract the ## Context section.
    * Returns { context: string } or { context: null } if not found.
-   * Falls back to legacy CONTEXT.md if PROJECT.md has no ## Context section.
+   * Reads ## Context section from PROJECT.md body.
    */
   async getProjectContext() {
     const content = await safeReadFile(path.join(this.planningDir, 'PROJECT.md'));
@@ -124,13 +117,6 @@ class PlanningReader {
       }
     }
 
-    // Backwards compat: read legacy CONTEXT.md
-    const legacyContext = await safeReadFile(path.join(this.planningDir, 'CONTEXT.md'));
-    if (legacyContext) {
-      const { body: ctxBody } = parseFrontmatter(legacyContext);
-      return { context: ctxBody.trim() || null };
-    }
-
     return { context: null };
   }
 
@@ -331,7 +317,7 @@ class PlanningReader {
         title: titleMatch ? titleMatch[1] : r.file.replace(/\.md$/, '').replace(/-/g, ' '),
         status: r.status || 'collected',
         source: r.mode || 'research',
-        date: r.Research_date || r.file.match(/^(\d{4}-\d{2}-\d{2})/)?.[1] || null,
+        date: r.research_date || r.file.match(/^(\d{4}-\d{2}-\d{2})/)?.[1] || null,
         relevance: r.confidence || 'medium',
         summary: r.body ? r.body.replace(/^#.+\n+/, '').slice(0, 300) : '',
         keyTakeaways: [],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sienklogic/plan-build-run",
-  "version": "2.19.1",
+  "version": "2.19.2",
   "description": "Plan it, Build it, Run it -- structured development workflow for Claude Code, OpenCode, Gemini and Codex",
   "bin": {
     "plan-build-run": "bin/install.js"

package/plan-build-run/bin/config-schema.json

@@ -288,7 +288,7 @@
         },
         "reverse_spec": {
           "type": "boolean",
-          "default": true,
+          "default": false,
           "description": "Enable reverse spec generation from existing code."
         },
         "predictive_impact": {
@@ -311,12 +311,6 @@
           "default": true,
           "description": "Enable team onboarding guide generation from ROADMAP.md, CONTEXT.md, and project conventions."
         },
-        "multi_layer_validation": {
-          "type": "boolean",
-          "default": false,
-          "deprecated": true,
-          "description": "DEPRECATED. Parallel multi-layer review passes were never integrated. Key retained for backward compatibility."
-        },
         "regression_prevention": {
           "type": "boolean",
           "default": true,
@@ -336,20 +330,25 @@
           "type": "boolean",
           "default": true,
           "description": "Enable architecture consistency guard that warns on dependency violations."
+        },
+        "cross_project_patterns": {
+          "type": "boolean",
+          "default": false,
+          "description": "Enable cross-project pattern detection and reuse across repositories."
+        },
+        "spec_templates": {
+          "type": "boolean",
+          "default": false,
+          "description": "Enable spec template generation from existing code patterns."
+        },
+        "global_learnings": {
+          "type": "boolean",
+          "default": false,
+          "description": "Enable global learnings aggregation across phases into KNOWLEDGE.md."
         }
       },
       "additionalProperties": false
     },
-    "validation_passes": {
-      "type": "array",
-      "items": {
-        "type": "string",
-        "enum": ["correctness", "security", "performance", "style", "tests", "accessibility", "docs", "deps"]
-      },
-      "default": ["correctness", "security"],
-      "deprecated": true,
-      "description": "DEPRECATED. Used only with multi_layer_validation which was never integrated. Key retained for backward compatibility."
-    },
     "autonomy": {
       "type": "object",
       "properties": {
@@ -1093,134 +1092,6 @@
       },
       "additionalProperties": false
     },
-    "local_llm": {
-      "type": "object",
-      "deprecated": true,
-      "description": "DEPRECATED. Local LLM infrastructure removed in v14.0. Key retained for backward compatibility.",
-      "properties": {
-        "enabled": {
-          "type": "boolean"
-        },
-        "provider": {
-          "type": "string",
-          "enum": [
-            "ollama"
-          ]
-        },
-        "endpoint": {
-          "type": "string",
-          "format": "uri"
-        },
-        "model": {
-          "type": "string"
-        },
-        "timeout_ms": {
-          "type": "integer",
-          "minimum": 500
-        },
-        "max_retries": {
-          "type": "integer",
-          "minimum": 0,
-          "maximum": 3
-        },
-        "fallback": {
-          "type": "string",
-          "enum": [
-            "frontier",
-            "skip"
-          ]
-        },
-        "routing_strategy": {
-          "type": "string",
-          "enum": [
-            "local_first",
-            "frontier_first"
-          ]
-        },
-        "features": {
-          "type": "object",
-          "properties": {
-            "artifact_classification": {
-              "type": "boolean"
-            },
-            "task_validation": {
-              "type": "boolean"
-            },
-            "plan_adequacy": {
-              "type": "boolean"
-            },
-            "gap_detection": {
-              "type": "boolean"
-            },
-            "context_summarization": {
-              "type": "boolean"
-            },
-            "source_scoring": {
-              "type": "boolean"
-            },
-            "commit_classification": {
-              "type": "boolean"
-            },
-            "test_triage": {
-              "type": "boolean"
-            },
-            "file_intent_classification": {
-              "type": "boolean"
-            }
-          },
-          "additionalProperties": false
-        },
-        "metrics": {
-          "type": "object",
-          "properties": {
-            "enabled": {
-              "type": "boolean"
-            },
-            "log_file": {
-              "type": "string"
-            },
-            "show_session_summary": {
-              "type": "boolean"
-            },
-            "frontier_token_rate": {
-              "type": "number",
-              "minimum": 0
-            }
-          },
-          "additionalProperties": false
-        },
-        "advanced": {
-          "type": "object",
-          "properties": {
-            "confidence_threshold": {
-              "type": "number",
-              "minimum": 0,
-              "maximum": 1
-            },
-            "max_input_tokens": {
-              "type": "integer",
-              "minimum": 100
-            },
-            "keep_alive": {
-              "type": "string"
-            },
-            "num_ctx": {
-              "type": "integer",
-              "minimum": 512
-            },
-            "disable_after_failures": {
-              "type": "integer",
-              "minimum": 1
-            },
-            "shadow_mode": {
-              "type": "boolean"
-            }
-          },
-          "additionalProperties": false
-        }
-      },
-      "additionalProperties": false
-    },
     "intel": {
       "type": "object",
       "properties": {
@@ -1297,6 +1168,13 @@
           "maximum": 1,
           "default": 1,
           "description": "Minimum completion confidence (0.5-1.0) required to skip verification. 1.0 means 100% must-haves must pass."
+        },
+        "min_full_percent": {
+          "type": "integer",
+          "minimum": 0,
+          "maximum": 100,
+          "default": 30,
+          "description": "Minimum percentage of phases in autonomous mode that must receive full verification (not just confidence-gate). 0 = all confidence-gate, 100 = all full verification. Default 30 means at least 30% of phases get full /pbr:review verification."
         }
       },
       "additionalProperties": false

package/plugins/pbr/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "pbr",
-  "version": "2.19.1",
+  "version": "2.19.2",
   "description": "Plan-Build-Run — Structured development workflow for Claude Code. Solves context rot through disciplined subagent delegation, structured planning, atomic execution, and goal-backward verification.",
   "author": {
     "name": "SienkLogic",

package/plugins/pbr/agents/advisor-researcher.md

@@ -9,6 +9,7 @@ tools:
   - WebSearch
   - WebFetch
 color: "#60A5FA"
+memory: none
 ---
 
 <files_to_read>

package/plugins/pbr/agents/debugger.md

@@ -196,10 +196,6 @@ When you need human input, emit a checkpoint block. Always include `Debug file:`
 | `HUMAN-ACTION` | User must do something you cannot | action needed, why, steps |
 | `DECISION` | Investigation branched | options with pros/cons, recommendation |
 
-## Local LLM Error Classification (DEPRECATED -- no-op)
-
-The local_llm feature was removed in v14.0. These commands are no-ops that return `{ deprecated: true }`. Skip this section and proceed with manual hypothesis generation.
-
 ## 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/pbr/agents/researcher.md

@@ -75,10 +75,6 @@ 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 (DEPRECATED -- no-op)
-
-The local_llm feature was removed in v14.0. These commands are no-ops that return `{ deprecated: true }`. Skip this section and assign source levels manually per the hierarchy table above.
-
 ---
 
 ## Confidence Levels

package/plugins/pbr/agents/synthesizer.md

@@ -201,10 +201,6 @@ 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 (DEPRECATED -- no-op)
-
-The local_llm feature was removed in v14.0. These commands are no-ops that return `{ deprecated: true }`. Skip this section and read documents normally.
-
 ## Context Budget
 
 ### Context Quality Tiers

package/plugins/pbr/dist/check-config-change.js

@@ -64,8 +64,6 @@ function validateConfig(configPath) {
     }
   }
 
-  // local_llm is deprecated — no advisory for missing key
-
   // Check version
   if (config.version && config.version < 2) {
     warnings.push(`Config version ${config.version} is outdated — expected version 2+`);
@@ -92,11 +90,6 @@ function validateConfig(configPath) {
     }
   }
 
-  // DEPRECATED: local_llm infrastructure removed in v14.0
-  if (config.local_llm && config.local_llm.enabled === true) {
-    warnings.push('local_llm feature is deprecated and has no effect. Set enabled: false to suppress this warning.');
-  }
-
   return warnings;
 }
 

package/plugins/pbr/dist/check-cross-plugin-sync.js

@@ -29,7 +29,7 @@ function checkCrossPluginSync(data) {
 
   let stagedFiles;
   try {
-    stagedFiles = execSync('git diff --cached --name-only', { encoding: 'utf8' }).trim().split('\n').filter(Boolean);
+    stagedFiles = execSync('git diff --cached --name-only', { encoding: 'utf8', stdio: ['pipe', 'pipe', 'pipe'] }).trim().split('\n').filter(Boolean);
   } catch (_e) {
     return null;
   }