@hanzlaa/rcode 4.7.3 → 4.9.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hanzlaa/rcode",
-  "version": "4.7.3",
+  "version": "4.9.0",
   "description": "rcode — the AI team that never forgets. Persistent memory, specialist agents, and slash commands for AI IDEs. Works in Claude Code, Cursor, Gemini, VS Code, and Antigravity.",
   "main": "cli/index.js",
   "bin": {

package/rcode/agents/rcode-code-reviewer.md

@@ -6,7 +6,7 @@ color: purple
 ---
 
 @.rcode/references/response-style.md
-@.rcode/references/karpathy-guidelines-full.md
+@.rcode/references/karpathy-guidelines.md
 @.rcode/references/no-unauthorized-git-ops.md
 @.rcode/references/code-reviewer-playbook.md
 

package/rcode/agents/rcode-docs-auditor.md

@@ -6,7 +6,7 @@ color: yellow
 ---
 
 @.rcode/references/response-style.md
-@.rcode/references/karpathy-guidelines-full.md
+@.rcode/references/karpathy-guidelines.md
 @.rcode/references/no-unauthorized-git-ops.md
 @.rcode/references/auditor-shared-checklists.md
 @.rcode/references/docs-auditor-playbook.md

package/rcode/agents/rcode-edge-case-hunter.md

@@ -6,7 +6,7 @@ color: red
 ---
 
 @.rcode/references/response-style.md
-@.rcode/references/karpathy-guidelines-full.md
+@.rcode/references/karpathy-guidelines.md
 @.rcode/references/no-unauthorized-git-ops.md
 @.rcode/references/auditor-shared-checklists.md
 

package/rcode/agents/rcode-haitham.md

@@ -12,7 +12,7 @@ description: |
   typography / colour system (use Zahra), architecture decisions (use Waleed),
   backend / API (use Yousef), test strategy (use Fatima), strategic priority
   (use Sadiq).
-tools: Read, Grep, Glob, Bash, WebFetch
+tools: Read, Grep, Glob, Bash, WebFetch, Write, Edit
 color: cyan
 ---
 

package/rcode/agents/rcode-hanzla.md

@@ -8,7 +8,7 @@ description: |
   Do NOT use for: architecture (Waleed), backend perf / queues (Yousef),
   frontend / RTL / a11y (Haitham), UX flows (Layla), test strategy (Fatima),
   deployment / CI (Khalid), scope (Hussain-PM).
-tools: Read, Grep, Glob, Bash
+tools: Read, Grep, Glob, Bash, Write, Edit
 color: green
 ---
 

package/rcode/agents/rcode-hussain-pm.md

@@ -9,7 +9,7 @@ description: |
   Do NOT use for: technical feasibility (Waleed), implementation (Hanzla /
   Yousef / Haitham), market positioning (Mariam), strategic go/no-go and
   kill criteria (Sadiq), QA test strategy (Fatima), sprint scrum ops (Hussain-SM).
-tools: Read, Grep, Glob, WebFetch
+tools: Read, Grep, Glob, WebFetch, Write, Edit
 color: orange
 ---
 

package/rcode/agents/rcode-nyquist-auditor.md

@@ -1,7 +1,7 @@
 ---
 name: rcode-nyquist-auditor
 description: Fills Nyquist validation gaps by generating tests and verifying coverage for phase requirements
-tools: Read, Grep, Glob, Bash
+tools: Read, Grep, Glob, Bash, Write, Edit
 color: purple
 ---
 

package/rcode/agents/rcode-omar.md

@@ -10,7 +10,7 @@ description: |
   deep frontend (use Haitham), deep backend perf (use Yousef), test strategy
   (use Fatima), scope / PRD (use Hussain-PM), strategic priority (use Sadiq),
   ML / RAG / embeddings (use Zayd), DevOps / deployment (use Khalid).
-tools: Read, Grep, Glob, Bash
+tools: Read, Grep, Glob, Bash, Write, Edit
 color: green
 ---
 

package/rcode/agents/rcode-remediation-planner.md

@@ -1,7 +1,7 @@
 ---
 name: rcode-remediation-planner
 description: Remediation Planner — spawned to plan remediation for issues, blockers, and failures. Creates action plans to recover from deviations, resolve blockers, and get back on track.
-tools: Read, Grep, Glob, Bash, Edit
+tools: Read, Grep, Glob, Bash
 color: orange
 ---
 

package/rcode/agents/rcode-security-adversary.md

@@ -6,7 +6,7 @@ color: red
 ---
 
 @.rcode/references/response-style.md
-@.rcode/references/karpathy-guidelines-full.md
+@.rcode/references/karpathy-guidelines.md
 @.rcode/references/no-unauthorized-git-ops.md
 @.rcode/references/auditor-shared-checklists.md
 

package/rcode/agents/rcode-security-auditor.md

@@ -6,7 +6,7 @@ color: purple
 ---
 
 @.rcode/references/response-style.md
-@.rcode/references/karpathy-guidelines-full.md
+@.rcode/references/karpathy-guidelines.md
 @.rcode/references/no-unauthorized-git-ops.md
 @.rcode/references/auditor-shared-checklists.md
 

package/rcode/agents/rcode-sprint-checker.md

@@ -6,7 +6,7 @@ color: green
 ---
 
 @.rcode/references/response-style.md
-@.rcode/references/karpathy-guidelines-full.md
+@.rcode/references/karpathy-guidelines.md
 @.rcode/references/sprint-checker-playbook.md
 
 <role>

package/rcode/agents/rcode-verifier.md

@@ -6,7 +6,7 @@ color: green
 ---
 
 @.rcode/references/response-style.md
-@.rcode/references/karpathy-guidelines-full.md
+@.rcode/references/karpathy-guidelines.md
 @.rcode/references/no-unauthorized-git-ops.md
 @.rcode/references/verifier-playbook.md
 

package/rcode/agents/rcode-waleed.md

@@ -10,7 +10,7 @@ description: |
   Do NOT use for: strategy / "should we build" (Sadiq), backend impl (Yousef),
   scope / PRD (Hussain-PM), test strategy (Fatima), market / GTM (Mariam),
   org-level multi-team coordination (Ahmed-Hassani-Director).
-tools: Read, Grep, Glob, Bash, WebFetch, WebSearch
+tools: Read, Grep, Glob, Bash, WebFetch, WebSearch, Write, Edit
 color: green
 ---
 

package/rcode/agents/rcode-yousef.md

@@ -11,7 +11,7 @@ description: |
   (use Haitham), test methodology (use Fatima), strategic priority (use Sadiq),
   scope / PRD (use Hussain-PM), implementation across the full stack
   (use Hanzla / Omar), deployment / CI (use Khalid).
-tools: Read, Grep, Glob, Bash, WebFetch
+tools: Read, Grep, Glob, Bash, WebFetch, Write, Edit
 color: blue
 ---
 

package/rcode/agents/rules/planner/task-templates.md

@@ -1,10 +1,12 @@
 # rcode Planner: Task Templates by Type
 
+<!-- id= and <title> are REQUIRED across every template below — this is what server/lib/scanner.js's buildPhaseTree actually parses. Do not use a `name` child tag or a bare title attribute; both are legacy/unsupported paths. -->
+
 ## Standard Task Template
 
 ```xml
-<task type="auto">
-  <name>Task: [Action-oriented name]</name>
+<task id="{sprint-id}.{NN}" type="auto">
+  <title>Task: [Action-oriented name]</title>
   <files>path/to/file.ext, path/to/another.ts</files>
   <action>[Specific implementation with what to avoid and WHY]</action>
   <verify>
@@ -21,8 +23,8 @@
 Use when you can write `expect(fn(input)).toBe(output)` before implementing.
 
 ```xml
-<task type="auto" tdd="true">
-  <name>Task: [name]</name>
+<task id="{sprint-id}.{NN}" type="auto" tdd="true">
+  <title>Task: [name]</title>
   <files>src/feature.ts, src/feature.test.ts</files>
   <behavior>
     - Test 1: [expected behavior]
@@ -44,8 +46,8 @@ Use when you can write `expect(fn(input)).toBe(output)` before implementing.
 Use for visual/functional verification (90% of checkpoints).
 
 ```xml
-<task type="checkpoint:human-verify" gate="blocking">
-  <name>Verify: [Feature name works]</name>
+<task id="{sprint-id}.{NN}" type="checkpoint:human-verify" gate="blocking">
+  <title>Verify: [Feature name works]</title>
   <what-built>[What the agent automated]</what-built>
   <how-to-verify>
     1. Navigate to http://localhost:3000/login
@@ -65,8 +67,8 @@ Use for visual/functional verification (90% of checkpoints).
 Use for implementation choices (9% of checkpoints).
 
 ```xml
-<task type="checkpoint:decision" gate="blocking">
-  <name>Decide: Database schema for messages</name>
+<task id="{sprint-id}.{NN}" type="checkpoint:decision" gate="blocking">
+  <title>Decide: Database schema for messages</title>
   <what-youre-deciding>
     Where to store user-to-message relationships:
     
@@ -94,8 +96,8 @@ Use for implementation choices (9% of checkpoints).
 Use for unavoidable manual steps (1% of checkpoints).
 
 ```xml
-<task type="checkpoint:human-action" gate="blocking">
-  <name>Set up Stripe API key</name>
+<task id="{sprint-id}.{NN}" type="checkpoint:human-action" gate="blocking">
+  <title>Set up Stripe API key</title>
   <what-needed>
     The next task implements Stripe payment processing, which requires your API key.
   </what-needed>
@@ -115,8 +117,8 @@ Use for unavoidable manual steps (1% of checkpoints).
 ## Database Migration Task Template
 
 ```xml
-<task type="auto">
-  <name>Task: Add role field to users table</name>
+<task id="{sprint-id}.{NN}" type="auto">
+  <title>Task: Add role field to users table</title>
   <files>prisma/schema.prisma, prisma/migrations/[timestamp]_add_role.sql</files>
   <action>
     1. Update User model in schema: add `role String @default("user")`
@@ -136,8 +138,8 @@ Use for unavoidable manual steps (1% of checkpoints).
 ## API Endpoint Task Template
 
 ```xml
-<task type="auto">
-  <name>Task: Implement POST /api/users endpoint</name>
+<task id="{sprint-id}.{NN}" type="auto">
+  <title>Task: Implement POST /api/users endpoint</title>
   <files>src/app/api/users/route.ts, src/types/user.ts</files>
   <action>
     Create POST endpoint accepting { email, password, name }:
@@ -167,8 +169,8 @@ Use for unavoidable manual steps (1% of checkpoints).
 ## UI Component Task Template
 
 ```xml
-<task type="auto">
-  <name>Task: Create LoginForm component</name>
+<task id="{sprint-id}.{NN}" type="auto">
+  <title>Task: Create LoginForm component</title>
   <files>src/components/LoginForm.tsx, src/components/LoginForm.test.tsx</files>
   <action>
     Create form component with:
@@ -195,8 +197,8 @@ Use for unavoidable manual steps (1% of checkpoints).
 ## Configuration Task Template
 
 ```xml
-<task type="auto">
-  <name>Task: Configure environment variables</name>
+<task id="{sprint-id}.{NN}" type="auto">
+  <title>Task: Configure environment variables</title>
   <files>.env.example, .env.local (local only)</files>
   <action>
     1. Create .env.example with template (no real values):
@@ -219,8 +221,8 @@ Use for unavoidable manual steps (1% of checkpoints).
 ## Documentation Task Template
 
 ```xml
-<task type="auto">
-  <name>Task: Document authentication flow</name>
+<task id="{sprint-id}.{NN}" type="auto">
+  <title>Task: Document authentication flow</title>
   <files>docs/AUTHENTICATION.md</files>
   <action>
     Document in docs/AUTHENTICATION.md:
@@ -242,8 +244,8 @@ Use for unavoidable manual steps (1% of checkpoints).
 ## Refactoring Task Template
 
 ```xml
-<task type="auto">
-  <name>Task: Extract auth validation to utility</name>
+<task id="{sprint-id}.{NN}" type="auto">
+  <title>Task: Extract auth validation to utility</title>
   <files>src/lib/auth-validation.ts, src/api/auth/login/route.ts</files>
   <action>
     1. Create src/lib/auth-validation.ts with functions:

package/rcode/agents/rules/verifier/verification-report.md

@@ -4,7 +4,7 @@
 
 **ALWAYS use the Write tool to create files** — never use `Bash(cat << 'EOF')` or heredoc commands for file creation.
 
-Create `.rcode/phases/{phase_dir}/{phase_num}-VERIFICATION.md`:
+Create `.planning/phases/{phase_dir}/{phase_num}-VERIFICATION.md`:
 
 ```markdown
 ---
@@ -108,7 +108,7 @@ Return with:
 
 **Status:** {passed | gaps_found | human_needed}
 **Score:** {N}/{M} must-haves verified
-**Report:** .rcode/phases/{phase_dir}/{phase_num}-VERIFICATION.md
+**Report:** .planning/phases/{phase_dir}/{phase_num}-VERIFICATION.md
 
 {If passed:}
 All must-haves verified. Phase goal achieved. Ready to proceed.

package/rcode/bin/lib/roadmap.cjs

@@ -326,7 +326,12 @@ function cmdUpdatePlanProgress(projectRoot, phaseNum, planId, status) {
   if (!phaseNum) {
     throw new Error('Usage: roadmap update-plan-progress <phase> [<plan-id> <status>]');
   }
-  // 1-arg form: auto-detect progress from disk
+  // 1-arg form: auto-detect progress from disk and persist it into the
+  // phase's **Status:**/**Plans:** prose block in ROADMAP.md.
+  // Fix #1013: this branch used to only *compute* status from disk and
+  // return it — it never read/wrote ROADMAP.md at all, so it always
+  // reported `updated: false` even when the computed status genuinely
+  // differed from the stale prose already in the file.
   if (!planId && !status) {
     const planningDir = path.join(projectRoot, '.planning', 'phases');
     let phaseDir = null;
@@ -344,12 +349,83 @@ function cmdUpdatePlanProgress(projectRoot, phaseNum, planId, status) {
     const planFiles = files.filter(f => /(?:^|-)(SPRINT|PLAN)\.md$/i.test(f));
     const summaryFiles = files.filter(f => /(?:^|-)SUMMARY\.md$/i.test(f));
     const allDone = planFiles.length > 0 && summaryFiles.length >= planFiles.length;
+    const computedStatus = allDone ? 'complete' : 'in_progress';
+    const note = `phase ${phaseNum}: ${planFiles.length} plan(s), ${summaryFiles.length} summary(s) — ${allDone ? 'complete' : 'in progress'}`;
+
+    const { path: rp } = resolveRoadmapPath(projectRoot);
+    if (!fs.existsSync(rp)) {
+      return { updated: false, note, plan_count: planFiles.length, summary_count: summaryFiles.length, status: computedStatus, error: 'ROADMAP.md not found' };
+    }
+    const content = fs.readFileSync(rp, 'utf8');
+    const phases = extractPhases(content);
+    const phase = phases.find((p) => normalizePhaseNum(p.number) === normalizePhaseNum(phaseNum));
+    if (!phase) {
+      return { updated: false, note, plan_count: planFiles.length, summary_count: summaryFiles.length, status: computedStatus, error: `phase ${phaseNum} not found in ROADMAP.md` };
+    }
+
+    const planIdRe = /^(\d+[A-Z]?(?:\.\d+)?-\d+)-(?:SPRINT|PLAN)\.md$/i;
+    const summaryIdRe = /^(\d+[A-Z]?(?:\.\d+)?-\d+)-SUMMARY\.md$/i;
+    const summarizedIds = new Set(
+      summaryFiles.map((f) => { const m = f.match(summaryIdRe); return m ? m[1] : null; }).filter(Boolean)
+    );
+    const planIds = [...new Set(
+      planFiles.map((f) => { const m = f.match(planIdRe); return m ? m[1] : null; }).filter(Boolean)
+    )].sort();
+
+    const section = content.slice(phase.headerIndex, phase.sectionEnd);
+    let newSection = section;
+    let changed = false;
+
+    // Update **Status:** — only when the computed status actually differs
+    // from what the prose says, so re-running this after completion is a
+    // no-op instead of rewriting the completion date every time.
+    const statusLineRe = /(\*\*Status(?::\*\*|\*\*:)\s*)([^\n]+)/i;
+    const statusMatch = section.match(statusLineRe);
+    if (statusMatch) {
+      const currentStatusNorm = statusMatch[2].trim().toLowerCase();
+      if (computedStatus === 'complete' && !currentStatusNorm.startsWith('complete')) {
+        const today = new Date().toISOString().slice(0, 10);
+        newSection = newSection.replace(statusLineRe, `$1Complete (${today})`);
+        changed = true;
+      } else if (computedStatus === 'in_progress' && currentStatusNorm.startsWith('planned')) {
+        newSection = newSection.replace(statusLineRe, `$1In Progress`);
+        changed = true;
+      }
+    }
+
+    // Replace a placeholder **Plans:** list (`- _TBD_` or empty) with the
+    // plan IDs actually found on disk. Leave already-populated lists alone.
+    if (planIds.length > 0) {
+      const plansBlockRe = /(\*\*Plans(?::\*\*|\*\*:)\s*\n)((?:\s*-\s*[^\n]+\n?)+)/i;
+      const plansMatch = newSection.match(plansBlockRe);
+      if (plansMatch) {
+        const isPlaceholder = /^\s*-\s*_TBD_\s*$/im.test(plansMatch[2]) || plansMatch[2].trim() === '';
+        if (isPlaceholder) {
+          const plansList = planIds
+            .map((id) => `- ${id} — ${summarizedIds.has(id) ? 'SUMMARY shipped' : 'in progress'}`)
+            .join('\n') + '\n';
+          newSection = newSection.replace(plansBlockRe, `$1${plansList}`);
+          changed = true;
+        }
+      }
+    }
+
+    if (!changed) {
+      return { updated: false, note: `${note} (ROADMAP.md already reflects this)`, plan_count: planFiles.length, summary_count: summaryFiles.length, status: computedStatus };
+    }
+
+    const newContent = content.slice(0, phase.headerIndex) + newSection + content.slice(phase.sectionEnd);
+    const tmp = rp + '.tmp';
+    fs.writeFileSync(tmp, newContent, 'utf8');
+    fs.renameSync(tmp, rp);
+
     return {
-      updated: false,
-      note: `phase ${phaseNum}: ${planFiles.length} plan(s), ${summaryFiles.length} summary(s) — ${allDone ? 'complete' : 'in progress'}`,
+      updated: true,
+      phase: phaseNum,
+      status: computedStatus,
       plan_count: planFiles.length,
       summary_count: summaryFiles.length,
-      status: allDone ? 'complete' : 'in_progress',
+      note: `ROADMAP.md phase ${phaseNum} block updated to ${computedStatus}`,
     };
   }
 

package/rcode/bin/rcode-tools.cjs

@@ -381,6 +381,7 @@ function cmdInit(workflowName, rawArgs) {
 
   let panel = [];
   let scores = {};
+  let domain = null;
 
   let agent_id = null;
 
@@ -395,8 +396,12 @@ function cmdInit(workflowName, rawArgs) {
     // Don't pad when user explicitly specified the agent list — their
     // choice is the final word.
     panel = filterPanelToInstalled(ideal, councilAgents, { pad: flags.agents.length === 0 });
+    // #1010 — domain must populate on every council session, not just
+    // --explain, because council.md's <output_format> banner prints
+    // Domain: unconditionally.
+    const explained = scorer.explainSelection(question, opts);
+    domain = explained.domain || null;
     if (flags.explain) {
-      const explained = scorer.explainSelection(question, opts);
       scores = explained.scores || {};
     }
   }
@@ -426,9 +431,14 @@ function cmdInit(workflowName, rawArgs) {
     flags,
     panel,
     scores,
+    domain,
     question_type: questionClassification.type,
     question_signals: questionClassification.signals,
     mode: config.mode || null,
+    // #1010 — response_language was only ever populated for phase-op/
+    // sprint-plan below; council.md documents it as a top-level init field
+    // and gates language pass-through to subagents on it being set.
+    response_language: config.response_language || config.language || null,
     config,
     installed_agents: installedAgents,
     paths: {
@@ -1519,6 +1529,10 @@ function cmdState(subArgs) {
   // =====================================================================
 
   // --- sprint add --phase NN --goal "Sprint goal" ---
+  // NOTE: this populates entry.sprints[] (an array). A separate code path,
+  // 'planned-phase' below (~line 3152), populates entry.plans (a plain count).
+  // These two fields are never reconciled with each other — see AUDIT-redundant-work.md
+  // finding 1 cross-check. Do not assume one implies the other is populated.
   if (sub === 'sprint' && subArgs[1] === 'add') {
     const flags = parseFlags(2);
     const state = readState() || defaultState();
@@ -3149,6 +3163,9 @@ function cmdState(subArgs) {
   // Execution-lifecycle phase state
   // =====================================================================
 
+  // NOTE: entry.plans (a count) is disjoint from entry.sprints[] (an array,
+  // set by 'sprint add' above) — see the comment there. Known schema divergence,
+  // not yet unified; do not read one as evidence the other is in sync.
   if (sub === 'planned-phase') {
     const flags = parseFlags(1);
     if (!flags.phase) throw new Error('planned-phase requires --phase <N>');
@@ -5385,6 +5402,109 @@ function cmdPlanCheckWaveOverlaps(rawArgs) {
   return { phase: phaseArg, phase_dir: path.relative(PROJECT_ROOT, phaseDir), plans_checked: plans.length, conflicts };
 }
 
+/**
+ * Deterministic frontend/backend glob check for classify-plan (issue #1021).
+ * Mirrors the FRONTEND_GLOBS / BACKEND_GLOBS rules formerly hand-applied by
+ * the orchestrating LLM in execute-waves.md — kept here as the single source
+ * of truth so both the CLI and the workflow doc describe the same behavior.
+ */
+function matchesFrontendGlob(file) {
+  const f = String(file || '').toLowerCase();
+  if (/\.(tsx|jsx|css)$/.test(f)) return true;
+  return f.includes('client') || f.includes('ui');
+}
+function matchesBackendGlob(file) {
+  const f = String(file || '').toLowerCase();
+  return f.includes('api') || f.includes('server') || f.includes('db') || f.includes('service');
+}
+
+const CLASSIFY_PLAN_ROUTE = { frontend: 'rcode-haitham', backend: 'rcode-yousef', 'full-stack': 'rcode-hanzla', other: 'rcode-executor' };
+
+function classifyPlanFiles(files, objective) {
+  const touchesFrontend = files.some(matchesFrontendGlob);
+  const touchesBackend = files.some(matchesBackendGlob);
+  let classification;
+  if (touchesFrontend && touchesBackend) classification = 'full-stack';
+  else if (touchesFrontend) classification = 'frontend';
+  else if (touchesBackend) classification = 'backend';
+  else classification = 'other';
+
+  if (classification === 'other') {
+    const obj = String(objective || '').toLowerCase();
+    const frontendKeywords = ['react', 'component', 'ui', 'css', 'tailwind', 'frontend', 'client-side', 'accessibility', 'a11y'];
+    const backendKeywords = ['api', 'endpoint', 'database', 'schema', 'service', 'queue', 'backend', 'server-side'];
+    if (frontendKeywords.some((k) => obj.includes(k))) classification = 'frontend';
+    else if (backendKeywords.some((k) => obj.includes(k))) classification = 'backend';
+  }
+  return classification;
+}
+
+/**
+ * classify-plan — deterministic replacement for execute-waves.md's
+ * hand-computed FRONTEND_GLOBS/BACKEND_GLOBS classification (issue #1021).
+ * A live execution run showed the orchestrating LLM never actually carried
+ * out the prose pseudocode, so a plan with a "db"-containing path still fell
+ * back to rcode-executor instead of rcode-yousef.
+ *
+ * Two call shapes:
+ *   classify-plan <phase> <plan-id>              — reads files_modified/objective from the plan's SPRINT.md
+ *   classify-plan --files=a,b,c --objective="..."  — classify an already-parsed list directly
+ */
+function cmdClassifyPlan(args) {
+  const flags = {};
+  const positional = [];
+  for (const t of args) {
+    if (t.startsWith('--files=')) flags.files = t.slice('--files='.length);
+    else if (t.startsWith('--objective=')) flags.objective = t.slice('--objective='.length);
+    else positional.push(t);
+  }
+
+  let files = [];
+  let objective = '';
+
+  if (flags.files !== undefined || flags.objective !== undefined) {
+    files = flags.files ? flags.files.split(',').map((s) => s.trim()).filter(Boolean) : [];
+    objective = flags.objective || '';
+  } else {
+    const [phaseArg, planArg] = positional;
+    if (!phaseArg || !planArg) {
+      throw new Error('Usage: classify-plan <phase> <plan-id>  OR  classify-plan --files=a,b,c --objective="text"');
+    }
+    const phasesDir = path.join(PLANNING_DIR, 'phases');
+    const norm = phaseArg.replace(/^0+/, '') || '0';
+    let phaseDir = null;
+    if (fs.existsSync(phasesDir)) {
+      for (const d of fs.readdirSync(phasesDir)) {
+        const m = d.match(/^(\d+)(?:[-.])/);
+        if (m && (m[1].replace(/^0+/, '') || '0') === norm) { phaseDir = path.join(phasesDir, d); break; }
+      }
+    }
+    if (!phaseDir) throw new Error(`Phase not found: ${phaseArg}`);
+    let planFile = null;
+    for (const file of fs.readdirSync(phaseDir).filter((f) => /-SPRINT\.md$/i.test(f)).sort()) {
+      const stem = file.replace(/-SPRINT\.md$/i, '');
+      if (stem === planArg || stem.endsWith(`-${planArg}`)) { planFile = file; break; }
+      const text = fs.readFileSync(path.join(phaseDir, file), 'utf8');
+      const { frontmatter } = parseFrontmatter(text);
+      if ((frontmatter.sprint || frontmatter.plan) === planArg) { planFile = file; break; }
+    }
+    if (!planFile) throw new Error(`Plan not found: ${planArg} in phase ${phaseArg}`);
+    const text = fs.readFileSync(path.join(phaseDir, planFile), 'utf8');
+    const { frontmatter, body } = parseFrontmatter(text);
+    let block = '';
+    if (text.startsWith('---\n')) {
+      const end = text.indexOf('\n---\n', 4);
+      if (end !== -1) block = text.slice(4, end);
+    }
+    files = fmListField(block, 'files_modified');
+    const objMatch = body.match(/^##\s+(?:Objective|Goal)\s*\n+([^\n]+)/mi);
+    objective = objMatch ? objMatch[1].trim() : (frontmatter.goal || '').replace(/^["']|["']$/g, '');
+  }
+
+  const classification = classifyPlanFiles(files, objective);
+  return { classification, subagent_type: CLASSIFY_PLAN_ROUTE[classification], files_checked: files.length };
+}
+
 /** phases list — directory inventory under .planning/phases with optional --type filter and --pick path. */
 function cmdPhasesList(args) {
   const argv = Array.isArray(args) ? args : String(args || '').trim().split(/\s+/).filter(Boolean);
@@ -6641,6 +6761,9 @@ async function main() {
       case 'phase-plan-index':
         result = cmdPhasePlanIndex(args.join(' '));
         break;
+      case 'classify-plan':
+        result = cmdClassifyPlan(args);
+        break;
       case 'phases':
         if (args[0] === 'list') { result = cmdPhasesList(args.slice(1)); if (result === undefined) return; }
         else { console.error('Unknown phases subcommand. Valid: list'); process.exit(1); }

package/rcode/brain/best-practices/state-sync-rule.md

@@ -8,7 +8,7 @@ Immediately after appending content to any of:
 
 - `.planning/ROADMAP.md` — milestones and phases
 - `.planning/epics.md` — epics and stories
-- `.rcode/phases/{phase}/sprint-{N}.md` — sprint commitments
+- `.planning/phases/{phase-dir}/{phase}-{plan}-SPRINT.md` — sprint commitments
 
 Call the state-sync helper:
 

package/rcode/references/council-protocol.md

@@ -32,7 +32,7 @@ Agreements without new information are silence. If an agent genuinely has nothin
 
 ## Panel selection (deterministic scoring)
 
-rcode's council uses a pure-function keyword scorer (`cli/lib/council-panel.cjs`) instead of LLM judgment. This is intentional:
+rcode's council uses a pure-function keyword scorer (`rcode/bin/lib/council-panel.cjs` source, installed at `.rcode/bin/lib/council-panel.cjs`) instead of LLM judgment. This is intentional:
 
 - **Deterministic:** same question produces same panel every time
 - **Testable:** the scoring function has unit tests
@@ -47,13 +47,18 @@ The scoring table is versioned with the package. Users who want different weight
 - Scope/feature questions (containing "scope", "feature", "roadmap", "prd") always include Hussain-PM.
 - If fewer than the minimum panel size (default 3) score non-zero, the panel is padded with the STRATEGIC_PADDING_ORDER list: sadiq, hussain-pm, waleed, fatima, nasser.
 
-**v2 prototype note:** only Sadiq, Waleed, and Fatima are installed as first-class subagents. The scorer may select agent ids that don't yet have subagent files — the orchestrator must filter to installed agents and fall back to the 3-agent panel.
+**Installed roster:** the full agent roster (currently 45 agents under `rcode/agents/`) is installed as first-class subagents. The scorer may still select an agent id that isn't present in a given project's installed set — the orchestrator filters the scored panel down to `installed_agents` from `init council` and pads from the fallback order if needed.
 
 ## Response presentation
 
-Subagent responses are presented **verbatim and in panel order**. The orchestrator never summarizes, paraphrases, or condenses agent output. The user came to the council to hear the agents speak in their own voices — summary defeats the point.
+Two presentation modes control how panel responses appear inline. The session artifact (see below) always stores the full verbatim text regardless of mode — only the inline presentation differs:
 
-After all verbatim responses, the orchestrator may add a single **Orchestrator Note** (max 3 sentences) that flags a disagreement worth following up on or recommends a second round. The Orchestrator Note is clearly labeled so it's not confused with agent speech.
+- **Default (compact summary)** — scannable in ~20 seconds: a one-sentence paraphrased position per agent, a convergence/divergence table, and Round 2 deltas. No verbatim transcripts inline.
+- **Verbose** (`--verbose` flag or `output.verbose: true` in config) — full verbatim responses printed in panel order, no summarization.
+
+The user came to the council to hear the agents' actual positions, not filler agreement — so even in compact mode, paraphrase the substance, not "I agree with X."
+
+After the panel output, the orchestrator may add a single **Orchestrator Note** (max 2-3 sentences) that flags a disagreement worth following up on or recommends a second round. The Orchestrator Note is clearly labeled so it's not confused with agent speech.
 
 ## Session artifact format
 

package/rcode/references/execute-auto-copy-learnings.md

@@ -0,0 +1,15 @@
+# execute.md — auto_copy_learnings (global learnings store)
+
+Extracted from `execute.md`'s `auto_copy_learnings` step. Only loaded when `features.global_learnings` is enabled (disabled by default) — see the conditional include at that point in `execute.md`.
+
+**Auto-copy phase learnings to global store (when enabled).**
+
+This step runs AFTER phase completion and SUMMARY.md is written. It copies any LEARNINGS.md
+entries from the completed phase to the global learnings store at `.rcode/knowledge/`.
+
+1. Check if LEARNINGS.md exists in the phase directory (use the `phase_dir` value from init context)
+2. If found, copy to global store:
+```bash
+node ".rcode/bin/rcode-tools.cjs" learnings copy 2>/dev/null || echo "⚠ Learnings copy failed — continuing"
+```
+Copy failure must NOT block phase completion.