@bonesofspring/ai-rules 0.2.26 → 0.2.28

package/presets/opencode/svelte/skills/feature-delivery/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: feature-delivery
-description: Delivers SvelteKit frontend features end-to-end across domain types, API services, store, UI, mocks, tests, and validation. Use when implementing or extending a feature, especially through /task or feature-developer.
+description: "Delivers SvelteKit frontend features end-to-end across domain types, API services, store, UI, mocks, tests, and validation. Use when implementing or extending a feature, especially through /task or feature-developer."
 ---
 
 # Feature Delivery

package/presets/opencode/svelte/skills/mockup-review/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: mockup-review
-description: Reviews a Figma mockup for design-system compliance, feature coverage, and viewport matrix via Figma MCP. Use for standalone mockup-review intent or when a feature/refactor/spike/spec-gen task includes a figma.com URL.
+description: "Reviews a Figma mockup for design-system compliance, feature coverage, and viewport matrix via Figma MCP. Use for standalone mockup-review intent or when a feature/refactor/spike/spec-gen task includes a figma.com URL."
 ---
 
 # Mockup review

package/presets/opencode/svelte/skills/playwright-e2e/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: playwright-e2e
-description: Plans, generates, and heals Playwright e2e tests using this preset's tests/e2e layout, *.cases.md plans, page objects, and shared helpers. Use for e2e planning, test generation, or fixing failing browser tests.
+description: "Plans, generates, and heals Playwright e2e tests using this preset's tests/e2e layout, *.cases.md plans, page objects, and shared helpers. Use for e2e planning, test generation, or fixing failing browser tests."
 ---
 
 # Playwright E2E

package/presets/opencode/svelte/skills/spec-authoring/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: spec-authoring
-description: Author or bootstrap consumer docs/specs (INDEX, page/feature templates, frontmatter v1). Use with task-analyst author mode, /spec-start, or when brief AC requests Generate/Bootstrap.
+description: "Author or bootstrap consumer docs/specs (INDEX, page/feature templates, frontmatter v1). Use with task-analyst author mode, /spec-start, or when brief AC requests Generate/Bootstrap."
 ---
 
 # Spec authoring

package/presets/opencode/svelte/skills/spec-discovery/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: spec-discovery
-description: Opt-in research-before-spec for task-analyst author mode. 11 named methods plus a mandatory claim ledger. Use when the user or AC asks discovery before generating docs/specs — not default /task consume.
+description: "Opt-in research-before-spec for task-analyst author mode. 11 named methods plus a mandatory claim ledger. Use when the user or AC asks discovery before generating docs/specs — not default /task consume."
 ---
 
 # Spec discovery

package/presets/opencode/svelte/skills/svelte-performance/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: svelte-performance
-description: SvelteKit performance checklist for implementers — bundle, lazy routes, dynamic import, load waterfalls, hydration, Core Web Vitals. Use while coding a svelte feature; not a substitute for the performance-auditor agent when the user asks for an audit report.
+description: "SvelteKit performance checklist for implementers — bundle, lazy routes, dynamic import, load waterfalls, hydration, Core Web Vitals. Use while coding a svelte feature; not a substitute for the performance-auditor agent when the user asks for an audit report."
 ---
 
 # Svelte performance (SvelteKit)

package/presets/opencode/svelte/skills/technical-retro/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: technical-retro
-description: Facilitates technical retrospectives for tasks, sprints, incidents, releases, and agent-team pipelines. Use when the user asks for retro, postmortem, process review, or action items after delivery.
+description: "Facilitates technical retrospectives for tasks, sprints, incidents, releases, and agent-team pipelines. Use when the user asks for retro, postmortem, process review, or action items after delivery."
 ---
 
 # Technical Retro

package/presets/opencode/svelte/skills/unit-testing/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: unit-testing
-description: Plans, generates, and heals unit tests for mappers, Svelte stores/runes actions, HTTP client behavior, and Svelte components using colocated *.spec.ts files and tests-unit rules. Use for unit test planning, generation, or fixing failing unit tests.
+description: "Plans, generates, and heals unit tests for mappers, Svelte stores/runes actions, HTTP client behavior, and Svelte components using colocated *.spec.ts files and tests-unit rules. Use for unit test planning, generation, or fixing failing unit tests."
 ---
 
 # Unit Testing

package/presets/opencode/svelte/skills/ux-design/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: ux-design
-description: Behavioral UX synthesis and review for task-analyst. Modes synthesize|review. Standalone intent ux-design or embedded required-signal / advisory UI AC. No Figma MCP; no DESIGN.md.
+description: "Behavioral UX synthesis and review for task-analyst. Modes synthesize|review. Standalone intent ux-design or embedded required-signal / advisory UI AC. No Figma MCP; no DESIGN.md."
 ---
 
 # UX design

package/presets/opencode/svelte/skills/write-adr/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: write-adr
-description: Author an Architecture Decision Record (ADR) with Context, Decision, Consequences, and Links. Use when capturing a durable technical choice for the team.
+description: "Author an Architecture Decision Record (ADR) with Context, Decision, Consequences, and Links. Use when capturing a durable technical choice for the team."
 ---
 
 # Write ADR

package/scripts/capture-cli-golden.mjs

@@ -28,11 +28,14 @@ const tmp = fs.mkdtempSync(path.join(os.tmpdir(), `ai-rules-golden-${tool}-`));
 try {
   runInit(cli, tmp, tool, preset);
   const files = snapshotInstall(tmp);
-  if (tool === 'mavis') {
-    delete files['.mavis/.ai-rules-install.json'];
-  }
-  if (tool === 'opencode') {
-    delete files['.opencode/.ai-rules-install.json'];
+  const manifestKey = {
+    cursor: '.cursor/.ai-rules-install.json',
+    claude: '.claude/.ai-rules-install.json',
+    mavis: '.mavis/.ai-rules-install.json',
+    opencode: '.opencode/.ai-rules-install.json',
+  }[tool];
+  if (manifestKey) {
+    delete files[manifestKey];
   }
   const golden = goldenDocument({
     tool,

package/scripts/check-claude-agent-blocks.mjs

@@ -0,0 +1,82 @@
+#!/usr/bin/env node
+/**
+ * Fail when a committed Claude agent «Rules to load» section differs from
+ * `expectedRulesToLoadBlock` (same helper as the inject codemod and the
+ * OpenCode/Mavis generators).
+ *
+ * Usage: node scripts/check-claude-agent-blocks.mjs [--stack <name>]
+ */
+import fs from 'node:fs';
+import path from 'node:path';
+import { fileURLToPath } from 'node:url';
+import {
+  expectedRulesToLoadBlock,
+  extractRulesToLoadBlock,
+  indexClaudeRuleStems,
+  loadReleasedStacks,
+  PLATFORM_RULES_INTRO,
+} from './lib/generate-adapter-common.mjs';
+import { getStackProfile, getZoneRulesForAgent, parseStackArg } from './lib/stack-platform-profiles.mjs';
+
+const scriptDirectory = path.dirname(fileURLToPath(import.meta.url));
+const packageRoot = path.join(scriptDirectory, '..');
+const presetsRoot = path.join(packageRoot, 'presets');
+
+function normalize(text) {
+  return text.replace(/\r\n/g, '\n').trim();
+}
+
+function checkStack(stack, mismatches) {
+  const profile = getStackProfile(stack);
+  const claudeRulesDir = path.join(presetsRoot, 'claude', stack, 'rules');
+  const { availableStems, pathForStem } = indexClaudeRuleStems(claudeRulesDir);
+  const agentsDir = path.join(presetsRoot, 'claude', stack, 'agents');
+  if (!fs.existsSync(agentsDir)) {
+    mismatches.push(`claude/${stack}: missing agents dir`);
+    return;
+  }
+  const agents = fs
+    .readdirSync(agentsDir)
+    .filter((entry) => entry.endsWith('.md') && entry !== 'README.md');
+  if (agents.length === 0) {
+    mismatches.push(`claude/${stack}: no agent files`);
+    return;
+  }
+  for (const entry of agents) {
+    const agentStem = entry.slice(0, -'.md'.length);
+    const text = fs.readFileSync(path.join(agentsDir, entry), 'utf8');
+    const expected = normalize(
+      expectedRulesToLoadBlock({
+        rulesPath: '.claude/rules',
+        alwaysOn: profile.alwaysOn,
+        zoneRules: getZoneRulesForAgent(stack, agentStem),
+        availableStems,
+        intro: PLATFORM_RULES_INTRO.claude,
+        pathForStem,
+      }),
+    );
+    const actual = normalize(extractRulesToLoadBlock(text));
+    if (actual !== expected) {
+      mismatches.push(`claude/${stack}/agents/${entry}`);
+    }
+  }
+}
+
+function main() {
+  const explicitStack = parseStackArg(process.argv.slice(2), { required: false });
+  const stacks = explicitStack ? [explicitStack] : loadReleasedStacks(presetsRoot, 'opencode');
+  const mismatches = [];
+  for (const stack of stacks) {
+    checkStack(stack, mismatches);
+  }
+  if (mismatches.length > 0) {
+    console.error(`FAIL: ${mismatches.length} Claude agent block(s) differ from expectedRulesToLoadBlock`);
+    for (const item of mismatches) {
+      console.error(`  ${item}`);
+    }
+    process.exit(1);
+  }
+  console.log(`OK: Claude agent blocks match expectedRulesToLoadBlock (${stacks.length} stacks)`);
+}
+
+main();

package/scripts/check-frontmatter.mjs

@@ -0,0 +1,83 @@
+#!/usr/bin/env node
+/**
+ * Frontmatter validity gate (dependency-free heuristic).
+ *
+ * Flags the YAML failure modes that shipped silently before: unquoted scalars
+ * containing `: ` / ` #`, and unquoted list items starting with a YAML-special
+ * character (`*`, `&`, `!`, `{`, `[`, `|`, `>`, `@`, `%`, backtick). Also flags
+ * tab indentation. Exit 1 on any finding.
+ *
+ * Full YAML parsing is intentionally avoided: the published package has zero
+ * runtime dependencies and the structure workflow does not install dev deps.
+ */
+import fs from 'node:fs';
+import path from 'node:path';
+import { fileURLToPath } from 'node:url';
+
+const scriptDirectory = path.dirname(fileURLToPath(import.meta.url));
+const packageDirectory = path.resolve(scriptDirectory, '..');
+const presetsRoot = path.join(packageDirectory, 'presets');
+
+const SCALAR_KEYS = /^([A-Za-z0-9_-]+):[ \t]*(.*)$/;
+const LIST_ITEM = /^([ \t]*-[ \t]+)(.*)$/;
+const SAFE_SCALAR = /^(['"]).*\1$|^(true|false|null|~)$|^-?\d+(\.\d+)?$|^[|>][-+]?$|^\[\]$|^\{\}$/;
+const SPECIAL_LEAD = /^[*&!%@`]/;
+
+const problems = [];
+
+function checkLine(file, lineNumber, line) {
+  if (/^\t|^[ \t]*\t/.test(line)) {
+    problems.push(`${file}:${lineNumber}: tab indentation is not valid YAML`);
+    return;
+  }
+  const kv = line.match(SCALAR_KEYS);
+  if (kv && kv[2] !== '' && !SAFE_SCALAR.test(kv[2])) {
+    const value = kv[2];
+    if (/: /.test(value) || /:[ \t]*$/.test(value) || /\s#/.test(value) || SPECIAL_LEAD.test(value)) {
+      problems.push(
+        `${file}:${lineNumber}: unquoted YAML value needs quoting: ${line.trim()}`,
+      );
+    }
+    return;
+  }
+  const item = line.match(LIST_ITEM);
+  if (item && !SAFE_SCALAR.test(item[2]) && SPECIAL_LEAD.test(item[2])) {
+    problems.push(`${file}:${lineNumber}: unquoted YAML list item needs quoting: ${line.trim()}`);
+  }
+}
+
+function walk(dir) {
+  for (const entry of fs.readdirSync(dir, { withFileTypes: true })) {
+    const full = path.join(dir, entry.name);
+    if (entry.isDirectory()) {
+      walk(full);
+    } else if (/\.(md|mdc)$/.test(entry.name)) {
+      const text = fs.readFileSync(full, 'utf8');
+      if (!text.startsWith('---\n')) {
+        continue;
+      }
+      const end = text.indexOf('\n---\n', 4);
+      if (end < 0) {
+        continue;
+      }
+      const frontmatter = text.slice(4, end);
+      const rel = path.relative(packageDirectory, full);
+      frontmatter.split('\n').forEach((line, index) => checkLine(rel, index + 2, line));
+    }
+  }
+}
+
+walk(presetsRoot);
+
+if (problems.length > 0) {
+  console.error(`Frontmatter check: ${problems.length} problem(s)`);
+  for (const problem of problems.slice(0, 50)) {
+    console.error(`  ${problem}`);
+  }
+  if (problems.length > 50) {
+    console.error(`  ... and ${problems.length - 50} more`);
+  }
+  process.exit(1);
+}
+
+console.log('Frontmatter check: OK');

package/scripts/check-generated-drift.mjs

@@ -0,0 +1,106 @@
+#!/usr/bin/env node
+/**
+ * Generated-tree drift gate.
+ *
+ * Regenerates `presets/mavis/<stack>` and `presets/opencode/<stack>` from the
+ * Cursor sources into a temp root and byte-compares them with the committed
+ * trees. Catches two classes of drift that the adapter-map checks cannot:
+ *   - hand-edits to a generated file;
+ *   - generator logic changes that were not followed by a regeneration.
+ *
+ * Exit 1 on any drift; prints the offending paths.
+ */
+import crypto from 'node:crypto';
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
+import { fileURLToPath } from 'node:url';
+import { generateMavisStack } from './lib/generate-mavis-adapter.mjs';
+import { generateOpencodeStack } from './lib/generate-opencode-adapter.mjs';
+import { loadReleasedStacks } from './lib/generate-adapter-common.mjs';
+
+const scriptDirectory = path.dirname(fileURLToPath(import.meta.url));
+const packageDirectory = path.resolve(scriptDirectory, '..');
+const presetsRoot = path.join(packageDirectory, 'presets');
+
+function walkRelative(root) {
+  const out = [];
+  const visit = (dir) => {
+    for (const entry of fs.readdirSync(dir, { withFileTypes: true })) {
+      const full = path.join(dir, entry.name);
+      if (entry.isDirectory()) {
+        visit(full);
+      } else if (entry.isFile()) {
+        out.push(path.relative(root, full).split(path.sep).join('/'));
+      }
+    }
+  };
+  if (fs.existsSync(root)) {
+    visit(root);
+  }
+  return out.sort();
+}
+
+function sha256File(filePath) {
+  return crypto.createHash('sha256').update(fs.readFileSync(filePath)).digest('hex');
+}
+
+function compareTrees(expectedRoot, actualRoot, label, drift) {
+  const expected = walkRelative(expectedRoot);
+  const actual = walkRelative(actualRoot);
+  const expectedSet = new Set(expected);
+  const actualSet = new Set(actual);
+  for (const rel of expected) {
+    if (!actualSet.has(rel)) {
+      drift.push(`${label}: committed file missing from generator output: ${rel}`);
+    } else if (sha256File(path.join(expectedRoot, rel)) !== sha256File(path.join(actualRoot, rel))) {
+      drift.push(`${label}: content drift in ${rel}`);
+    }
+  }
+  for (const rel of actual) {
+    if (!expectedSet.has(rel)) {
+      drift.push(`${label}: generator output has an uncommitted file: ${rel}`);
+    }
+  }
+}
+
+const tempRoot = fs.mkdtempSync(path.join(os.tmpdir(), 'ai-rules-generated-drift-'));
+const drift = [];
+try {
+  fs.mkdirSync(path.join(tempRoot, 'presets'), { recursive: true });
+  fs.cpSync(path.join(presetsRoot, 'cursor'), path.join(tempRoot, 'presets', 'cursor'), {
+    recursive: true,
+  });
+  fs.cpSync(path.join(presetsRoot, '_shared'), path.join(tempRoot, 'presets', '_shared'), {
+    recursive: true,
+  });
+
+  for (const [platform, generate] of [
+    ['mavis', generateMavisStack],
+    ['opencode', generateOpencodeStack],
+  ]) {
+    const stacks = loadReleasedStacks(presetsRoot, platform);
+    for (const stack of stacks) {
+      generate(stack, { root: tempRoot });
+      compareTrees(
+        path.join(presetsRoot, platform, stack),
+        path.join(tempRoot, 'presets', platform, stack),
+        `${platform}/${stack}`,
+        drift,
+      );
+    }
+  }
+} finally {
+  fs.rmSync(tempRoot, { recursive: true, force: true });
+}
+
+if (drift.length > 0) {
+  console.error(`Generated-tree drift detected (${drift.length} issue(s)):`);
+  for (const line of drift) {
+    console.error(`  ${line}`);
+  }
+  console.error('Run: node scripts/generate-mavis.mjs --stack <stack> and generate-opencode.mjs --stack <stack>');
+  process.exit(1);
+}
+
+console.log('Generated-tree drift check: OK (mavis + opencode match generator output)');

package/scripts/check-mavis-token-budget.mjs

@@ -10,7 +10,7 @@ import { loadReleasedMavisStacks } from './lib/discover-hooks.mjs';
 
 const scriptDirectory = path.dirname(fileURLToPath(import.meta.url));
 const presetsRoot = path.join(scriptDirectory, '..', 'presets');
-const ROOT_ENTRY_CAP = 4;
+const ROOT_ENTRY_CAP = 5;
 
 const stacks = loadReleasedMavisStacks(presetsRoot);
 if (stacks.length === 0) {

package/scripts/check-opencode-token-budget.mjs

@@ -10,7 +10,7 @@ import { loadReleasedOpencodeStacks } from './lib/discover-hooks.mjs';
 
 const scriptDirectory = path.dirname(fileURLToPath(import.meta.url));
 const presetsRoot = path.join(scriptDirectory, '..', 'presets');
-const ROOT_ENTRY_CAP = 4;
+const ROOT_ENTRY_CAP = 5;
 
 const stacks = loadReleasedOpencodeStacks(presetsRoot);
 if (stacks.length === 0) {

package/scripts/check-preset-structure.sh

@@ -47,6 +47,24 @@ else
 fi
 echo
 
+echo ">>> generated-drift"
+if node "$SCRIPT_DIR/check-generated-drift.mjs"; then
+  echo "<<< generated-drift: OK"
+else
+  echo "<<< generated-drift: FAIL"
+  status=1
+fi
+echo
+
+echo ">>> claude-agent-blocks"
+if node "$SCRIPT_DIR/check-claude-agent-blocks.mjs"; then
+  echo "<<< claude-agent-blocks: OK"
+else
+  echo "<<< claude-agent-blocks: FAIL"
+  status=1
+fi
+echo
+
 echo ">>> opencode-token-budget"
 if node "$SCRIPT_DIR/check-opencode-token-budget.mjs"; then
   echo "<<< opencode-token-budget: OK"
@@ -138,6 +156,24 @@ run_gate "chain-hook-coverage" "$SCRIPT_DIR/check-chain-hook-coverage.sh"
 run_gate "task-router-intents" "$SCRIPT_DIR/check-task-router-intents.sh"
 run_gate "i18n" "$SCRIPT_DIR/check-preset-i18n.sh" AI_RULES_I18N_FAIL=1 AI_RULES_I18N_SOFT=0
 
+echo ">>> frontmatter"
+if node "$SCRIPT_DIR/check-frontmatter.mjs"; then
+  echo "<<< frontmatter: OK"
+else
+  echo "<<< frontmatter: FAIL"
+  status=1
+fi
+echo
+
+echo ">>> probe-mavis-capabilities"
+if node "$SCRIPT_DIR/test-probe-mavis-capabilities.mjs"; then
+  echo "<<< probe-mavis-capabilities: OK"
+else
+  echo "<<< probe-mavis-capabilities: FAIL"
+  status=1
+fi
+echo
+
 echo ">>> cli-mcp-merge"
 if node "$SCRIPT_DIR/test-cli-mcp-merge.mjs"; then
   echo "<<< cli-mcp-merge: OK"

package/scripts/check-preset-token-budget.sh

@@ -16,8 +16,10 @@ from pathlib import Path
 
 presets = Path(os.environ["AI_RULES_PRESETS_ROOT"])
 STACKS = ("next", "nuxt", "svelte", "ios-swift", "android-kotlin", "go", "java", "mcp-ts", "php-hexagonal", "php-laravel")
-BODY_CAP = 140  # applied to 3-rule trios (Claude session-start, Cursor trio when len==3)
-BODY_CAP_CURSOR_4 = 170  # Cursor now allows 4 always-on (next-app-core, package-manager, code-quality, ai-local-overlay)
+MIN_ALWAYS_ON = 3  # base invariants every stack must keep session-start
+MAX_ALWAYS_ON = 5  # aligned cap for Cursor / Claude / Mavis / OpenCode always-on rules
+BODY_CAP = 200  # session-start body-sum cap (Cursor and Claude)
+BODY_CAP_STRETCH = 170  # informational stretch target
 # Slim-companion rules (UI-edit bundle, slim app-cores, slim tooling) may be
 # shorter than 15 lines **if and only if** their stem is in the allow-list and
 # body is <= MIN_SLIM_BODY_LINES. A slim rule whose body grows past this cap
@@ -134,7 +136,7 @@ failed = 0
 
 print("Preset token / session-start budget (Option A)")
 print(f"Presets: {presets}")
-print(f"Body line cap (Claude trio): {BODY_CAP}; (Cursor 4-set): {BODY_CAP_CURSOR_4}")
+print(f"Always-on count: {MIN_ALWAYS_ON}..{MAX_ALWAYS_ON}; body cap: {BODY_CAP} (stretch {BODY_CAP_STRETCH})")
 print()
 
 for stack in STACKS:
@@ -151,14 +153,13 @@ for stack in STACKS:
             print(f"FAIL: meta preset-* must not be alwaysApply: {p.name}")
             failed += 1
 
-    if len(always) not in (3, 4):
-        print(f"FAIL: Cursor alwaysApply count is {len(always)}, expected 3 or 4 (4th slot reserved for ai-local-overlay)")
+    if not (MIN_ALWAYS_ON <= len(always) <= MAX_ALWAYS_ON):
+        print(f"FAIL: Cursor alwaysApply count is {len(always)}, expected {MIN_ALWAYS_ON}..{MAX_ALWAYS_ON}")
         failed += 1
-    elif len(always) == 4:
-        # 4-set: ensure the 4th rule is ai-local-overlay (or declares itself as such).
-        extra = [p for p in always if p.stem not in {"code-quality-and-refactoring", "package-manager", "next-app-core", "nuxt-app-core", "svelte-app-core", "ios-app-core", "android-app-core", "go-app-core", "java-app-core", "mcp-ts-app-core", "php-app-core", "laravel-app-core", "architecture-boundaries", "mcp-server-boundaries"}]
-        if not any(p.stem == "ai-local-overlay" for p in always) and extra:
-            print(f"FAIL: Cursor alwaysApply=4 but no ai-local-overlay.mdc; the 4th slot is reserved for the local overlay rule")
+    elif len(always) > 3:
+        # Slots beyond the base invariants are reserved for ai-local-overlay.
+        if not any(p.stem == "ai-local-overlay" for p in always):
+            print(f"FAIL: Cursor alwaysApply={len(always)} but no ai-local-overlay.mdc; extra slots are reserved for the local overlay rule")
             failed += 1
 
     # --- Claude no-paths (excl README) ---
@@ -171,19 +172,17 @@ for stack in STACKS:
             print(f"FAIL: meta preset-* must not be session-start: {rel}")
             failed += 1
 
-    if len(session) != 3:
+    if not (MIN_ALWAYS_ON <= len(session) <= MAX_ALWAYS_ON):
         print(
             f"FAIL: Claude session-start count is {len(session)}, "
-            "expected exactly 3 (Option A excludes **/README.md)"
+            f"expected {MIN_ALWAYS_ON}..{MAX_ALWAYS_ON} (Option A excludes **/README.md)"
         )
         failed += 1
 
     # --- Body sum of intended session-start set (Cursor alwaysApply + Claude session) ---
-    # Cursor allows 3 or 4 rules (4th slot reserved for ai-local-overlay); Claude is exactly 3.
+    # Cursor and Claude allow 3..5 always-on rules; extra slots reserved for ai-local-overlay.
     for label, files in (("Cursor", always), ("Claude", session)):
-        if label == "Cursor" and len(files) not in (3, 4):
-            continue
-        if label == "Claude" and len(files) != 3:
+        if not (MIN_ALWAYS_ON <= len(files) <= MAX_ALWAYS_ON):
             continue
         total = 0
         for path in files:
@@ -193,7 +192,7 @@ for stack in STACKS:
             total += n
             name = path.name if label == "Cursor" else path.relative_to(claude_rules).as_posix()
             print(f"  {label} body lines {name}: {n}")
-        cap = BODY_CAP_CURSOR_4 if label == "Cursor" and len(always) == 4 else BODY_CAP
+        cap = BODY_CAP
         print(f"  {label} session-start body sum: {total} (cap {cap}, files={len(files)})")
         if total > cap:
             print(

package/scripts/check-shared-core-drift.sh

@@ -80,15 +80,20 @@ def consumers(stem: str) -> list[Path]:
                     out.append(path)
             except OSError:
                 continue
-    contract_path = presets / "mavis" / "platform-contract.json"
-    if contract_path.is_file():
+    # Generated platform trees are driven by their own contracts. Scan every
+    # released stack for both mavis and opencode so a core change that is not
+    # regenerated cannot slip through unnoticed.
+    for platform in ("mavis", "opencode"):
+        contract_path = presets / platform / "platform-contract.json"
+        if not contract_path.is_file():
+            continue
         try:
             contract = json.loads(contract_path.read_text(encoding="utf-8"))
             for stack in contract.get("releasedStacks") or []:
-                mavis_base = presets / "mavis" / stack
-                if not mavis_base.is_dir():
+                platform_base = presets / platform / stack
+                if not platform_base.is_dir():
                     continue
-                for path in mavis_base.rglob("*"):
+                for path in platform_base.rglob("*"):
                     if path.suffix not in {".md", ".mdc"}:
                         continue
                     try:

package/scripts/close-agent-task-human-gate.mjs

@@ -4,7 +4,7 @@
  * Prefer invoking the hook directly from /task-continue (see command Metrics section).
  *
  * Usage:
- *   node close-agent-task-human-gate.mjs [--cwd <repo>] [--platform cursor|claude|mavis|opencode] [--slug <slug>]
+ *   node close-agent-task-human-gate.mjs [--cwd <repo>] [--platform cursor|claude|mavis|opencode] [--slug <slug>] [--no-activate-next]
  */
 
 import { execFileSync } from 'node:child_process';
@@ -14,7 +14,7 @@ import path from 'node:path';
 const PLATFORMS = new Set(['cursor', 'claude', 'mavis', 'opencode']);
 
 function parseArgs(argv) {
-  const options = { cwd: process.cwd(), platform: 'cursor', slug: null };
+  const options = { cwd: process.cwd(), platform: 'cursor', slug: null, activateNext: true };
   function nextValue(flag, i) {
     const next = argv[i + 1];
     if (next === undefined || next.startsWith('-')) {
@@ -35,7 +35,9 @@ function parseArgs(argv) {
       options.slug = nextValue(arg, i);
       i += 1;
     } else if (arg === '--activate-next') {
-      /* always activate via hook payload */
+      options.activateNext = true;
+    } else if (arg === '--no-activate-next') {
+      options.activateNext = false;
     } else if (arg === '--team-root') {
       const teamRoot = path.resolve(nextValue(arg, i));
       options.cwd = path.resolve(teamRoot, '..', '..');
@@ -76,7 +78,7 @@ if (options.slug) {
 const payload = JSON.stringify({
   status: 'completed',
   action: 'end_human_gate',
-  activateNext: true,
+  activateNext: options.activateNext,
 });
 
 const output = execFileSync('bash', [hookPath], {