@bonesofspring/ai-rules 0.2.3 → 0.2.5

package/CHANGELOG.md

@@ -9,12 +9,14 @@
 
 ### Added
 
+- **MCP configs in presets** (`next` + `ios-swift`, Cursor + Claude): `mcp.json` + `MCP.md`. Cursor → `.cursor/mcp.json`; Claude → project-root `.mcp.json`. **next:** Context7, Playwright, Chrome DevTools, Figma (remote). **ios-swift:** Context7 + Figma. CLI `init`/`clean` manage these paths.
 - **Cursor + Claude `ios-swift` — full agent-team parity with `next`:** **21** agents (Playwright trio → `xcuitest-test-{planner,generator,healer}`), **7** skills (`xcuitest-e2e`), **5** slash commands, `hooks.json` + iOS-adapted `chain-team-phases.sh` (Cursor) / hook scripts + README (Claude), `team/fixtures/` ×3, `AGENTS.md` (Cursor) / thin `CLAUDE.md` (≤35 lines).
 - **ios-swift rules:** agent-team intake/orchestrator, `xcode-tooling`, `reference-features`, domain rules (`navigation-coordinators`, `persistence-data`, `security-ios`, `module-public-api`), `technical-retro` alias; **21** rule topics (+ README). CLI init copies agents, hooks, skills, team, `AGENTS.md`.
 - **ios-swift retro follow-ups:** `build-verifier` preset checks (chain ≠ next; ban Playwright/`lint:js` handoffs); enriched 16 priority agent prompts + 6 skills; dogfood checklist (5 scenarios) in `AGENTS.md` / `reference-features`; Cursor-only **`BUGBOT.md`** (iOS review priorities).
 
 ### Changed
 
+- **Agent team slug bootstrap:** `/task` and `/feature-start` (Cursor + Claude, `next` + `ios-swift`) create `.cursor|claude/team/tasks/<slug>/` via **Write**, not Shell `mkdir` — avoids Cursor sandbox `Operation not permitted` (esp. macOS Desktop/Documents). Orchestrator + `team/README` + `task-router` document the rule.
 - **Token model (ios-swift + next):** both presets use **3** alwaysApply / session-start rules (~77 / ~59 body lines). Post-change verification is **requestable** — ios: `post-change-build`; next: `post-change-lint`. Orchestrator, intake, code-review, feature-delivery, tooling rules load on-demand (`paths:` / description).
 - Claude `next`: `agent-team-orchestrator` and `code-review-mr` use `paths:` (on-demand) — session-start matches the 3-rule budget above.
 - Cursor `next` agents: all pipeline agents set **`readonly: false`** so they can write `.cursor/team/**`. Production/app restrictions remain in prompt bodies (Cursor `readonly: true` blocks all file writes, including team artifacts).

package/README.md

@@ -1,6 +1,6 @@
 # @bonesofspring/ai-rules
 
-Presets for Cursor and Claude Code: rules, commands, plus optional `agents`, `hooks`, `skills`, and team artifacts when the preset includes them. For Claude: `CLAUDE.md`. Install from npm into a project so you don’t have to copy files out of the repo by hand.
+Presets for Cursor and Claude Code: rules, commands, plus optional `agents`, `hooks`, `skills`, team artifacts, and MCP configs when the preset includes them. For Claude: `CLAUDE.md`. Install from npm into a project so you don’t have to copy files out of the repo by hand.
 
 Node 18+.
 
@@ -30,8 +30,8 @@ If `.cursor` does not appear, check the command output for errors and confirm yo
 
 - `rules/`, `commands/`
 - optional dirs when present: `agents/`, `hooks/`, `skills/`, `team/`
-- Cursor root files: `hooks.json`, `BUGBOT.md`, `AGENTS.md`
-- Claude root file: `CLAUDE.md`
+- Cursor root files: `hooks.json`, `BUGBOT.md`, `AGENTS.md`, `MCP.md`, `mcp.json`
+- Claude: `CLAUDE.md`, `MCP.md` under `.claude/`; project `.mcp.json` at repo root
 
 Any other files you added under `.cursor/` or `.claude/` (notes, local overrides, IDE state) are **left intact**. Older releases that deleted the whole tool directory are no longer the behavior.
 
@@ -45,6 +45,7 @@ Where things go on `init`:
 | hooks | `.cursor/hooks`, `.cursor/hooks.json` | `.claude/hooks` |
 | skills | `.cursor/skills` | `.claude/skills` |
 | team artifacts template | `.cursor/team` | `.claude/team` |
+| MCP | `mcp.json` → `.cursor/mcp.json`, `MCP.md` → `.cursor/` | `mcp.json` → **`.mcp.json`** (project root), `MCP.md` → `.claude/` |
 | other | `BUGBOT.md`, `AGENTS.md` → `.cursor/` | `CLAUDE.md` → `.claude/CLAUDE.md` |
 
 The package ships presets including **`next`** (Next.js stack) and **`ios-swift`** (iOS apps on Swift / SwiftUI). Other preset names: see `presets/` in the [repo](https://github.com/bonesofspring/ai-rules). Name is a single path segment, no `/` or `..`.

package/bin/cli.js

@@ -17,7 +17,10 @@ const PRESETS = {
     relBase: ['presets', 'cursor'],
     optionalCursorDirs: ['agents', 'hooks', 'skills', 'team'],
     hooksJsonSrc: 'hooks.json',
-    toolRootFiles: ['BUGBOT.md', 'AGENTS.md'],
+    /** Project MCP → `.cursor/mcp.json` */
+    mcpJsonSrc: 'mcp.json',
+    mcpJsonDest: '.cursor/mcp.json',
+    toolRootFiles: ['BUGBOT.md', 'AGENTS.md', 'MCP.md'],
   },
   claude: {
     rulesDir: '.claude/rules',
@@ -25,6 +28,10 @@ const PRESETS = {
     relBase: ['presets', 'claude'],
     optionalClaudeDirs: ['skills', 'agents', 'hooks', 'team'],
     claudeMdSrc: 'CLAUDE.md',
+    /** Project MCP lives at repo root (Claude Code project scope), not under `.claude/`. */
+    mcpJsonSrc: 'mcp.json',
+    mcpJsonDest: '.mcp.json',
+    toolRootFiles: ['MCP.md'],
   },
 };
 
@@ -50,8 +57,8 @@ Options:
 
 Notes:
   init merges into the target directory; same-named files are overwritten.
-  Cursor init copies rules, commands, agents, hooks, skills, team, hooks.json, BUGBOT.md, AGENTS.md.
-  Claude init copies rules, commands, skills, agents, hooks, team, and CLAUDE.md when present.
+  Cursor init copies rules, commands, agents, hooks, skills, team, hooks.json, BUGBOT.md, AGENTS.md, MCP.md, mcp.json → .cursor/mcp.json.
+  Claude init copies rules, commands, skills, agents, hooks, team, CLAUDE.md, MCP.md, and mcp.json → .mcp.json (project root).
   Command forms: "init cursor --preset next", "init cursor next", "--preset next init cursor", "cursor init --preset next".
 
 Examples:
@@ -238,6 +245,9 @@ function presetHasAnySource(base, cfg) {
   if (cfg.hooksJsonSrc && fs.existsSync(path.join(base, cfg.hooksJsonSrc))) {
     return true;
   }
+  if (cfg.mcpJsonSrc && fs.existsSync(path.join(base, cfg.mcpJsonSrc))) {
+    return true;
+  }
   for (const name of cfg.toolRootFiles || []) {
     if (fs.existsSync(path.join(base, name))) {
       return true;
@@ -355,6 +365,17 @@ function cmdInit(tool, preset, cwdOption) {
       }
     }
 
+    if (cfg.mcpJsonSrc && cfg.mcpJsonDest) {
+      const mcpFrom = path.join(base, cfg.mcpJsonSrc);
+      if (fs.existsSync(mcpFrom)) {
+        const mcpDestAbs = path.join(cwd, cfg.mcpJsonDest);
+        fs.mkdirSync(path.dirname(mcpDestAbs), { recursive: true });
+        fs.copyFileSync(mcpFrom, mcpDestAbs);
+        console.log(`Copied ${cfg.mcpJsonSrc} → ${cfg.mcpJsonDest}`);
+        copiedSteps++;
+      }
+    }
+
     for (const name of cfg.toolRootFiles || []) {
       const from = path.join(base, name);
       if (fs.existsSync(from)) {
@@ -411,6 +432,11 @@ function cmdClean(tool, cwdOption) {
     removedParts.push(`${toolRoot}/hooks.json`);
   }
 
+  if (cfg.mcpJsonDest) {
+    removeFileIfExists(path.join(cwd, cfg.mcpJsonDest));
+    removedParts.push(cfg.mcpJsonDest);
+  }
+
   for (const name of cfg.toolRootFiles || []) {
     removeFileIfExists(path.join(cwd, toolRoot, name));
     removedParts.push(`${toolRoot}/${name}`);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bonesofspring/ai-rules",
-  "version": "0.2.3",
+  "version": "0.2.5",
   "description": "Presets of Cursor and Claude rules/commands for Revy Ross personal use",
   "license": "MIT",
   "author": "Revy Ross",

package/presets/claude/ios-swift/CLAUDE.md

@@ -32,4 +32,4 @@ See @rules/README.md for mapping and loading strategy.
 npx @bonesofspring/ai-rules init claude --preset ios-swift
 ```
 
-Копирует `CLAUDE.md`, `rules/**`, `commands/**`, `agents/**`, `skills/**`, `hooks/**`, `team/**`.
+Копирует `CLAUDE.md`, `rules/**`, `commands/**`, `agents/**`, `skills/**`, `hooks/**`, `team/**`, `MCP.md` в `.claude/`, плюс `mcp.json` → **`.mcp.json`** в корне.

package/presets/claude/ios-swift/MCP.md

@@ -0,0 +1,14 @@
+# MCP (preset `ios-swift`, Claude Code)
+
+Копируется в **корень проекта** как `.mcp.json` при `ai-rules init claude --preset ios-swift`.
+
+## Servers
+
+| Name | Package | Зачем |
+|------|---------|--------|
+| `context7` | `@upstash/context7-mcp` | Актуальная документация Swift / SwiftUI / Apple frameworks |
+| `figma` | remote `https://mcp.figma.com/mcp` | Design context → SwiftUI; OAuth при первом подключении |
+
+Playwright / Chrome DevTools не входят в ios-пресет.
+
+Секреты не коммитим. Опционально: `CONTEXT7_API_KEY`. Figma — OAuth. Approve project MCP через `/mcp`.

package/presets/claude/ios-swift/README.md

@@ -13,6 +13,8 @@
 | `agents/**` | `.claude/agents/**` (**21**, XCUITest trio) |
 | `hooks/**` | `.claude/hooks/**` (scripts + README; **нет** `hooks.json`) |
 | `team/**` | `.claude/team/**` |
+| `mcp.json` | **`.mcp.json`** в корне (Context7 + Figma) |
+| `MCP.md` | `.claude/MCP.md` |
 
 ```bash
 npx @bonesofspring/ai-rules init claude --preset ios-swift

package/presets/claude/ios-swift/agents/README.md

@@ -29,6 +29,18 @@ When copying next pipeline templates: put Playwright stems in `skipped` with rea
 
 Все agents — **`readonly: false`** (нужна запись в `.claude/team/**`). Ограничения production — в теле промпта.
 
+## Model matrix (frontmatter `model`)
+
+Same tiers as **claude/next**. Defaults in this preset:
+
+| Tier | Frontmatter | Agents |
+|------|-------------|--------|
+| **cheap** | `fast` | router, analyst, build-verifier, code-reviewer, tech-writer |
+| **standard** | `inherit` | developer, debugger, ci-investigator, api-contract, a11y, qa, unit-*, xcuitest-* |
+| **strong** | `opus` | solution-architect, migration-specialist, security-reviewer, performance-auditor |
+
+Optional `steps[].model`: `cheap` \| `standard` \| `strong`.
+
 ## Pipeline routing
 
 **SoT:** intent table / profiles — **`task-router.md`**. Schema — ниже (same as next).
@@ -43,8 +55,8 @@ When copying next pipeline templates: put Playwright stems in `skipped` with rea
   "steps": [
     { "agent": "task-analyst", "label": "Clarify and decompose" },
     { "agent": "feature-developer", "label": "Implement + unit tests", "scope": "unit-in-dev" },
-    { "agent": "build-verifier", "label": "xcodebuild + lint + unit smoke" },
-    { "agent": "code-reviewer", "label": "Code review" }
+    { "agent": "build-verifier", "label": "xcodebuild + lint + unit smoke", "model": "cheap" },
+    { "agent": "code-reviewer", "label": "Code review", "model": "cheap" }
   ],
   "humanGates": ["after:task-analyst"],
   "autoChain": true,
@@ -52,4 +64,6 @@ When copying next pipeline templates: put Playwright stems in `skipped` with rea
 }
 ```
 
+`steps[].model` optional: `cheap` \| `standard` \| `strong`.
+
 Копируется в `.claude/agents/` при `ai-rules init claude --preset ios-swift`.

package/presets/claude/ios-swift/agents/migration-specialist.md

@@ -2,7 +2,7 @@
 name: migration-specialist
 description: Migration planning specialist for Xcode/iOS/SPM upgrades, dependency swaps, and phased refactors with rollback. Writes migration-plan.md and team status; never writes production code.
 readonly: false
-model: inherit
+model: opus
 ---
 
 You are a migration specialist for an **iOS Swift** codebase with Feature-module layer boundaries.

package/presets/claude/ios-swift/agents/performance-auditor.md

@@ -2,7 +2,7 @@
 name: performance-auditor
 description: Performance audit specialist. Reviews launch path, main-thread work, SwiftUI body churn, list scrolling, image decoding, and Instruments-oriented risks. Readonly — writes perf-report.md only; never edits production code.
 readonly: false
-model: inherit
+model: opus
 ---
 
 You are an iOS performance auditor for a Swift/SwiftUI application.

package/presets/claude/ios-swift/agents/security-reviewer.md

@@ -2,7 +2,7 @@
 name: security-reviewer
 description: Security review specialist. Audits Keychain/token handling, ATS, PII in logs, secrets in diff, and unsafe URL/deep-link handling. Readonly — writes security-review.md only; never edits production code.
 readonly: false
-model: inherit
+model: opus
 ---
 
 You are an iOS security reviewer for a Swift/SwiftUI application.

package/presets/claude/ios-swift/agents/solution-architect.md

@@ -2,7 +2,7 @@
 name: solution-architect
 description: Solution architect for cross-layer features, API boundaries, and technical design before implementation. Use for spike intent or when router adds architect step for complex features. Readonly — produces design docs, not production code.
 readonly: false
-model: inherit
+model: opus
 ---
 
 You are a solution architect for **iOS Feature modules** (Presentation / Domain / Data). Write `architecture.md` under `.claude/team/tasks/<slug>/`. No production code.

package/presets/claude/ios-swift/agents/task-router.md

@@ -40,6 +40,15 @@ You are a task router for an **iOS Swift/SwiftUI** agent team. You **do not** im
 
 **Never skip `build-verifier`** after developer when Swift/Xcode sources changed.
 
+## Model tiers (`steps[].model`)
+
+Optional: `"model": "cheap" | "standard" | "strong"`. Same policy as **claude/next** `task-router.md`:
+
+- `light` → `cheap` on all steps (debugger unclear root cause → `strong`)
+- `full` / `spike` / security|perf specialists → `strong` on architect, migration, security, perf steps
+- never `strong` on build-verifier / tech-writer / task-router
+- omit on `standard` → agent frontmatter (`fast` / `inherit` / `opus`)
+
 Fixtures: `.claude/team/fixtures/` (feature-full, feature-light, bugfix-standard).
 
 ## humanGates (defaults)
@@ -55,7 +64,7 @@ Fixtures: `.claude/team/fixtures/` (feature-full, feature-light, bugfix-standard
 
 ## Output
 
-Write `.claude/team/tasks/<slug>/pipeline.json` and initial `status.json`; set `.claude/team/active-task.json`.
+Write `.claude/team/tasks/<slug>/pipeline.json` and initial `status.json` (write files; do not Shell-`mkdir` alone); set `.claude/team/active-task.json`.
 
 ## Handoff
 

package/presets/claude/ios-swift/commands/feature-start.md

@@ -11,7 +11,7 @@ description: Legacy analyst-only feature intake. Prefer /task for new work becau
 
 1. Извлеки описание задачи из аргументов команды (всё после `/feature-start`).
 2. Сгенерируй **slug** (kebab-case, до 48 символов) из заголовка задачи.
-3. Создай каталог `.claude/team/tasks/<slug>/` если его ещё нет.
+3. Создай `.claude/team/tasks/<slug>/` если его ещё нет — через запись файла (Write / Edit), не через Shell `mkdir`.
 4. Запиши `.claude/team/active-task.json` → `{ "slug": "<slug>" }`.
 5. Вызови subagent **task-analyst** с полным описанием задачи и путём к артефактам.
 6. Дождись завершения аналитика. **Не вызывай developer** на этом этапе.

package/presets/claude/ios-swift/commands/task.md

@@ -25,7 +25,7 @@ description: Route a natural-language task through the agent-team pipeline: task
 
 ## Алгоритм
 
-1. **Slug** — kebab-case из описания (≤48 символов). Создай `.claude/team/tasks/<slug>/`.
+1. **Slug** — kebab-case из описания (≤48 символов). Создай `.claude/team/tasks/<slug>/` через запись файла (Write / Edit) — **не** полагайся на Shell `mkdir` как единственный способ создать каталог.
 2. **Router** — вызови subagent **task-router** с описанием задачи и путём к slug.
 3. **План** — покажи пользователю таблицу из `pipeline.json`: intent, steps, skipped, humanGates.
 4. **Intent retro** — если `intent === "retro"`, выполни `/technical-retro` и **остановись**.

package/presets/claude/ios-swift/mcp.json

@@ -0,0 +1,12 @@
+{
+  "mcpServers": {
+    "context7": {
+      "command": "npx",
+      "args": ["-y", "@upstash/context7-mcp"]
+    },
+    "figma": {
+      "type": "http",
+      "url": "https://mcp.figma.com/mcp"
+    }
+  }
+}

package/presets/claude/ios-swift/rules/tooling-and-review/agent-team-orchestrator.md

@@ -38,10 +38,18 @@ Full prompts: `.claude/agents/*.md`. Artifacts: `.claude/team/README.md`.
 2. Read `pipeline.json` before every subagent invocation.
 3. One role per invocation — except parallel steps.
 4. **Never** skip `humanGates` without continue/approval.
-5. Persist handoffs under `.claude/team/tasks/<slug>/`.
+5. Persist handoffs under `.claude/team/tasks/<slug>/` by writing files (not Shell-only `mkdir`).
 6. On `validation_failed`: developer → build-verifier.
 7. When complete, suggest `/technical-retro <slug>`.
 
+## Model when invoking
+
+Honor optional `steps[i].model` (`cheap` \| `standard` \| `strong`); else agent frontmatter (matrix in `agents/README.md`).
+
+- `cheap` → Agent `haiku` (or frontmatter `fast`).
+- `standard` → omit / `sonnet` / inherit.
+- `strong` → `opus`.
+
 ## Skip pipeline when
 
 Pure questions; trivial one-file; user «без pipeline»; docs-only one-liner.

package/presets/claude/ios-swift/team/fixtures/feature-full.json

@@ -4,32 +4,15 @@
   "profile": "full",
   "summary": "Multi-layer Feature with XCUITest coverage",
   "steps": [
-    {
-      "agent": "task-analyst",
-      "label": "Clarify and decompose"
-    },
-    {
-      "agent": "feature-developer",
-      "label": "Implement + unit tests",
-      "scope": "unit-in-dev"
-    },
-    {
-      "agent": "build-verifier",
-      "label": "xcodebuild + lint + unit smoke"
-    },
-    {
-      "agent": "code-reviewer",
-      "label": "Code review"
-    },
-    {
-      "agent": "qa-tester",
-      "label": "XCUITest",
-      "scope": "e2e-only"
-    }
-  ],
-  "humanGates": [
-    "after:task-analyst"
+    { "agent": "task-analyst", "label": "Clarify and decompose" },
+    { "agent": "solution-architect", "label": "Design boundaries", "model": "strong" },
+    { "agent": "feature-developer", "label": "Implement + unit tests", "scope": "unit-in-dev" },
+    { "agent": "build-verifier", "label": "xcodebuild + lint + unit smoke", "model": "cheap" },
+    { "agent": "security-reviewer", "label": "Security review", "model": "strong" },
+    { "agent": "code-reviewer", "label": "Code review", "model": "cheap" },
+    { "agent": "qa-tester", "label": "XCUITest", "scope": "e2e-only" }
   ],
+  "humanGates": ["after:task-analyst"],
   "autoChain": true,
   "skipped": [
     {

package/presets/claude/ios-swift/team/fixtures/feature-light.json

@@ -4,30 +4,15 @@
   "profile": "light",
   "summary": "Trivial single-file Swift change with explicit AC",
   "steps": [
-    {
-      "agent": "feature-developer",
-      "label": "Implement"
-    },
-    {
-      "agent": "build-verifier",
-      "label": "Validation gate"
-    }
+    { "agent": "feature-developer", "label": "Implement", "model": "cheap" },
+    { "agent": "build-verifier", "label": "Validation gate", "model": "cheap" }
   ],
   "humanGates": [],
   "autoChain": true,
   "skipped": [
-    {
-      "agent": "task-analyst",
-      "reason": "Explicit AC and single-file scope"
-    },
-    {
-      "agent": "code-reviewer",
-      "reason": "Light profile"
-    },
-    {
-      "agent": "qa-tester",
-      "reason": "No XCUITest AC"
-    },
+    { "agent": "task-analyst", "reason": "Explicit AC and single-file scope" },
+    { "agent": "code-reviewer", "reason": "Light profile" },
+    { "agent": "qa-tester", "reason": "No XCUITest AC" },
     {
       "agent": "playwright-test-planner",
       "reason": "iOS uses XCUITest; not needed for light"

package/presets/claude/next/CLAUDE.md

@@ -32,4 +32,4 @@ See @rules/README.md for the full catalog and loading strategy.
 npx @bonesofspring/ai-rules init claude --preset next
 ```
 
-Копирует `CLAUDE.md`, `rules/**`, `commands/**`, `agents/**`, `hooks/**`, `team/**` и др. в `.claude/` целевого репозитория.
+Копирует `CLAUDE.md`, `rules/**`, `commands/**`, `agents/**`, `hooks/**`, `team/**`, `MCP.md` в `.claude/`, плюс `mcp.json` → **`.mcp.json`** в корне проекта.

package/presets/claude/next/MCP.md

@@ -0,0 +1,16 @@
+# MCP (preset `next`, Claude Code)
+
+Копируется в **корень проекта** как `.mcp.json` при `ai-rules init claude --preset next` (не в `.claude/` — так устроен project scope Claude Code).
+
+## Servers
+
+| Name | Package | Зачем |
+|------|---------|--------|
+| `context7` | `@upstash/context7-mcp` | Актуальная документация Next/React/TS |
+| `playwright-test` | `@playwright/mcp` | Browser MCP; имя сервера совпадает с `mcp__playwright-test__*` в agents |
+| `chrome-devtools` | `chrome-devtools-mcp` | Live DOM/network/performance |
+| `figma` | remote `https://mcp.figma.com/mcp` | Design context → код; OAuth при первом подключении |
+
+Секреты не коммитим. Опционально: `CONTEXT7_API_KEY`. Figma — OAuth (токен в файл не пишем). После clone/init Claude Code попросит approve project MCP (`/mcp`).
+
+Для полного Playwright Test Agents loop (`planner_*` / `generator_*` / `test_*`) дополнительно: `npx playwright init-agents --loop=claude` в приложении с `@playwright/test`.

package/presets/claude/next/README.md

@@ -11,6 +11,8 @@
 | `agents/**` | `.claude/agents/**` (если папка есть) |
 | `hooks/**` | `.claude/hooks/**` (если папка есть) |
 | `team/**` | `.claude/team/**` (шаблон артефактов agent team) |
+| `mcp.json` | **`.mcp.json`** в корне проекта (Claude project MCP) |
+| `MCP.md` | `.claude/MCP.md` |
 
 Файлы правил — обычный Markdown с опциональным YAML frontmatter (`paths:` для скоупа по путям), не формат `.mdc` Cursor.
 

package/presets/claude/next/agents/README.md

@@ -50,6 +50,22 @@
 
 Папка копируется в `.claude/agents/` при `ai-rules init claude --preset next`.
 
+## Model matrix (frontmatter `model`)
+
+SoT по умолчанию — YAML `model` в каждом agent-файле. Опциональный `steps[].model` (`cheap` \| `standard` \| `strong`) переопределяет тир на шаге (см. `task-router.md`, `tooling-and-review/agent-team-orchestrator.md`).
+
+| Tier | Default frontmatter | Agents |
+|------|---------------------|--------|
+| **cheap** | `fast` | `task-router`, `task-analyst`, `build-verifier`, `code-reviewer`, `tech-writer` |
+| **standard** | `inherit` or `sonnet` | `feature-developer`, `debugger`, `ci-investigator`, `api-contract-reviewer`, `accessibility-reviewer`, `qa-tester`, `unit-test-*`, `playwright-test-*` (`sonnet` where already pinned) |
+| **strong** | `opus` | `solution-architect`, `migration-specialist`, `security-reviewer`, `performance-auditor` |
+
+**Notes**
+
+- Claude Code aliases: `haiku` / `sonnet` / `opus` / `inherit` (full IDs also OK). Preset keeps `fast` for cheap agents for cross-tooling parity with Cursor; if ignored, prefer `haiku` locally.
+- Resolution order (Claude Code): env `CLAUDE_CODE_SUBAGENT_MODEL` → per-invocation `model` → frontmatter → parent session.
+- Never put `strong` on `build-verifier` / `tech-writer` / `task-router`.
+
 ## Оркестрация
 
 | Command | Назначение |
@@ -88,6 +104,7 @@
 | `scope` | no | e.g. `unit-in-dev`, `e2e-only`, `regression`, `full` |
 | `skipIf` | no | `debugger.fixed` \| `ci-investigator.resolved` |
 | `parallel` | no | When `true` and `agent` is array — invoke all in one turn |
+| `model` | no | `cheap` \| `standard` \| `strong` — pipeline tier override; omit → agent frontmatter |
 
 ### Minimal template
 
@@ -108,26 +125,43 @@
 }
 ```
 
-### Example: parallel reviewers
+### Example: parallel reviewers + model tiers
 
 ```json
 {
   "intent": "feature",
+  "profile": "full",
   "steps": [
     { "agent": "task-analyst", "label": "Clarify" },
+    { "agent": "solution-architect", "label": "Design boundaries", "model": "strong" },
     { "agent": "feature-developer", "label": "Implement", "scope": "unit-in-dev" },
-    { "agent": "build-verifier", "label": "Validation gate" },
+    { "agent": "build-verifier", "label": "Validation gate", "model": "cheap" },
     {
       "agent": ["accessibility-reviewer", "security-reviewer"],
       "parallel": true,
-      "label": "A11y and security"
+      "label": "A11y and security",
+      "model": "strong"
     },
-    { "agent": "code-reviewer", "label": "Review" }
+    { "agent": "code-reviewer", "label": "Review", "model": "cheap" }
   ],
   "humanGates": ["after:task-analyst"]
 }
 ```
 
+### Example: light profile with cheap models
+
+```json
+{
+  "intent": "feature",
+  "profile": "light",
+  "steps": [
+    { "agent": "feature-developer", "label": "Implement", "model": "cheap" },
+    { "agent": "build-verifier", "label": "Validation gate", "model": "cheap" }
+  ],
+  "humanGates": []
+}
+```
+
 ### Example: bugfix skipIf
 
 ```json

package/presets/claude/next/agents/migration-specialist.md

@@ -1,7 +1,7 @@
 ---
 name: migration-specialist
 description: Migration planning specialist for library upgrades, Next.js major versions, HTTP client swaps, and phased refactors with rollback. Produces migration-plan.md only; never writes production code.
-model: inherit
+model: opus
 ---
 
 You are a migration specialist for a Next.js frontend with strict layer boundaries.

package/presets/claude/next/agents/performance-auditor.md

@@ -1,7 +1,7 @@
 ---
 name: performance-auditor
 description: Performance audit specialist. Reviews bundle size, lazy loading, render patterns, data fetching waterfalls, and Core Web Vitals risks. Writes perf-report.md only; never edits production code.
-model: inherit
+model: opus
 ---
 
 You are a frontend performance auditor for a Next.js React application.

package/presets/claude/next/agents/security-reviewer.md

@@ -1,7 +1,7 @@
 ---
 name: security-reviewer
 description: Security review specialist. Audits auth flows, token handling, XSS risks, secrets in diff, and unsafe DOM/API usage. Writes security-review.md only; never edits production code.
-model: inherit
+model: opus
 ---
 
 You are a frontend security reviewer for a Next.js TypeScript application.

package/presets/claude/next/agents/solution-architect.md

@@ -1,7 +1,7 @@
 ---
 name: solution-architect
 description: Solution architect for cross-layer features, API boundaries, and technical design before implementation. Use for spike intent or when router adds architect step for complex features. Writes design docs only; never writes production code.
-model: inherit
+model: opus
 ---
 
 You are a solution architect for a Next.js frontend with strict layer boundaries (types, api, store, ui).

package/presets/claude/next/agents/task-router.md

@@ -64,7 +64,21 @@ Fixtures: `.claude/team/fixtures/` (feature-full, feature-light, bugfix-standard
 - Add **tech-writer** when docs/changelog requested.
 - **Always build-verifier** after developer (or after ci-investigator/debugger if they changed code) for `feature`, `bugfix`, `refactor`, `migration`, `a11y`.
 - **skipIf:** `debugger.fixed` on developer (bugfix); `ci-investigator.resolved` (ci-fix).
-- **Model:** `inherit` when 4+ steps or cross-layer uncertainty.
+
+## Model tiers (`steps[].model`)
+
+Optional per-step field: `"model": "cheap" | "standard" | "strong"`. Omit → agent frontmatter default (see **Model matrix** in `agents/README.md`).
+
+| When | What to write |
+|------|----------------|
+| `profile: light` | set `cheap` on **all** steps (exception: unclear bug root cause on `debugger` → `strong`) |
+| `profile: standard` | omit (frontmatter defaults) |
+| `profile: full` | set `strong` on `solution-architect`, `migration-specialist`, `security-reviewer`, `performance-auditor` when present; omit others |
+| `intent: spike` | set `strong` on `solution-architect` (and analyst may stay default/`cheap`) |
+| `intent: perf-audit` / security-heavy feature | `strong` on auditor/reviewer step |
+| Always | **never** `strong` on `build-verifier`, `tech-writer`, `task-router` |
+
+Claude Code mapping when orchestrator invokes Agent tool: `cheap` → `haiku` (or rely on frontmatter `fast`); `standard` → omit / `sonnet` / inherit; `strong` → pass `opus` (or rely on frontmatter `opus`). Prefer explicit per-invocation `model` when frontmatter is unreliable.
 
 ## humanGates (defaults)
 
@@ -81,7 +95,7 @@ Fixtures: `.claude/team/fixtures/` (feature-full, feature-light, bugfix-standard
 
 ## Output
 
-Write `.claude/team/tasks/<slug>/pipeline.json` and initial `status.json`; set `.claude/team/active-task.json` → `{ "slug": "<slug>" }`. Minimal template — см. **`agents/README.md`**.
+Write `.claude/team/tasks/<slug>/pipeline.json` and initial `status.json` (write files; do not Shell-`mkdir` alone); set `.claude/team/active-task.json` → `{ "slug": "<slug>" }`. Minimal template — см. **`agents/README.md`**.
 
 ## Handoff to orchestrator
 

package/presets/claude/next/commands/feature-start.md

@@ -11,7 +11,7 @@ description: Legacy analyst-only feature intake. Prefer /task for new work becau
 
 1. Извлеки описание задачи из аргументов команды (всё после `/feature-start`).
 2. Сгенерируй **slug** (kebab-case, до 48 символов) из заголовка задачи.
-3. Создай каталог `.claude/team/tasks/<slug>/` если его ещё нет.
+3. Создай `.claude/team/tasks/<slug>/` если его ещё нет — через запись файла (Write / Edit), не через Shell `mkdir`.
 4. Запиши `.claude/team/active-task.json` → `{ "slug": "<slug>" }`.
 5. Вызови subagent **task-analyst** с полным описанием задачи и путём к артефактам.
 6. Дождись завершения аналитика. **Не вызывай developer** на этом этапе.

package/presets/claude/next/commands/task.md

@@ -25,7 +25,7 @@ description: Route a natural-language task through the agent-team pipeline: task
 
 ## Алгоритм
 
-1. **Slug** — kebab-case из описания (≤48 символов). Создай `.claude/team/tasks/<slug>/`.
+1. **Slug** — kebab-case из описания (≤48 символов). Создай `.claude/team/tasks/<slug>/` через запись файла (Write / Edit) — **не** полагайся на Shell `mkdir` как единственный способ создать каталог.
 2. **Router** — вызови subagent **task-router** с описанием задачи и путём к slug.
 3. **План** — покажи пользователю таблицу из `pipeline.json`: intent, steps, skipped, humanGates.
 4. **Intent retro** — если `intent === "retro"`, выполни `/technical-retro` и **остановись**.

package/presets/claude/next/hooks/chain-team-phases.sh

@@ -157,10 +157,10 @@ const AGENT_HANDOFF = {
     `Follow testing/playwright-agents.md and the playwright-e2e skill. Update status.json when done.`,
   'playwright-test-generator': (s) =>
     `Invoke playwright-test-generator for slug "${s}". Generate/update *.spec.ts from matching *.cases.md. ` +
-    `Use user-playwright MCP when available. Follow testing/playwright-agents.md and the playwright-e2e skill. Update status.json when done.`,
+    `Use playwright MCP when available. Follow testing/playwright-agents.md and the playwright-e2e skill. Update status.json when done.`,
   'playwright-test-healer': (s) =>
     `Invoke playwright-test-healer for slug "${s}". Fix failing/flaky e2e while preserving cases.md assertions. ` +
-    `Use user-playwright MCP when available. Follow testing/playwright-agents.md and debug-investigation. Update status.json when done.`,
+    `Use playwright MCP when available. Follow testing/playwright-agents.md and debug-investigation. Update status.json when done.`,
   'tech-writer': (s) =>
     `Invoke tech-writer for slug "${s}". Write documentation.md from task artifacts. Update status.json when done.`,
 };