@hecer/yoke 1.10.0 → 1.12.0

package/.claude-plugin/plugin.json

@@ -2,12 +2,12 @@
   "$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json",
   "name": "yoke",
   "displayName": "Yoke",
-  "version": "1.10.0",
+  "version": "1.12.0",
   "description": "Cross-agent coding harness: one curated skill canon (TDD, brainstorming, plans, reviews, shipping, design verification) plus mechanical safety gates and an autonomous loop via the yoke CLI.",
   "author": { "name": "HECer", "url": "https://github.com/HECer" },
   "homepage": "https://github.com/HECer/yoke#readme",
   "repository": "https://github.com/HECer/yoke",
   "license": "MIT",
-  "keywords": ["harness", "cross-agent", "tdd", "code-review", "autonomous-loop", "codex", "gemini-cli"],
+  "keywords": ["harness", "cross-agent", "tdd", "code-review", "autonomous-loop", "codex", "gemini-cli", "qwen-code"],
   "skills": "./canon/skills/"
 }

package/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "yoke",
-  "version": "1.10.0",
+  "version": "1.12.0",
   "description": "Cross-agent coding discipline, mechanical gates, and release workflows",
   "skills": "./canon/skills/",
   "hooks": "./hooks/hooks.json"

package/CHANGELOG.md

@@ -1,5 +1,42 @@
 # Changelog
 
+## 1.12.0 — 2026-09-08
+
+### Fixed
+- Parse Qwen Code's native assistant/result/structured-result output and cumulative native model statistics, ignoring nested subagent results and refusing terminal-error verdicts.
+- Use Qwen's native headless approval flags, retain the sandbox in safe mode and explicitly permit shell tools there; exclude native delegation tools when Yoke owns concurrency.
+- Detect native Qwen session/project markers and include Qwen in automatic loop review selection and CLI review validation.
+- Install manual Qwen skills with native invocation restrictions and a supported RTK PreToolUse retry guard; back up settings and remove only Yoke's obsolete BeforeTool hook on retrofit.
+- Accept up to 32 routing profiles so an all-agent setup remains valid.
+
+### Added
+- Opt-in `setup --model-provider=deepseek,kimi` with DeepSeek V4 Flash/Pro and Kimi K2.6/K2.7 Code/K3 API configurations, environment key references, preserved custom settings and routing profiles.
+- Explicit Qwen `PROTOCOL::MODEL` selectors mapped to native authentication/model arguments without changing global login settings.
+
+### Migration and validation limits
+- Fresh Qwen setups now use one standard profile with the user's configured model. Existing workers remain unchanged unless reset with `--routing-preset`; configure stronger profiles for stronger assessments.
+- API presets require separately configured credentials and are not model-quality or cost benchmarks. Read [Qwen model support](docs/QWEN-MODEL-SUPPORT.md) for exact behavior and migration.
+- Tested with regression fixtures and real Qwen Code 0.23.0 against a synthetic local tool-calling server; no authenticated provider benchmarks or production sandbox validation. This release targets npm package 1.12.0; publication is triggered by the published GitHub release and verified separately.
+
+## 1.11.0 — 2026-09-07
+
+### Added
+- Add Qwen Code (Alibaba) as fourth supported provider alongside Claude, Codex and Gemini.
+- Detect Qwen host environment via `QWEN_CLI` and `QWEN_CLI_HOME` environment variables.
+- Add Qwen routing workers with four capability tiers: `qwen-turbo-latest` (light), `qwen3-coder-plus` (standard/strong), `qwen3-235b-a22b` (frontier).
+- Parse Qwen streaming telemetry for token usage and model reporting.
+- Support Qwen in all CLI commands: `setup`, `loop`, `review`, `prd draft`, `prd assess`, `goal run`.
+
+### Changed
+- Update project description from "three agents" to "four agents" to reflect Qwen support.
+- Extend review resolution order to include Qwen for cross-model reviews.
+- Update setup prompts to offer Qwen as agent and runner option.
+
+### Migration and validation limits
+- Existing configurations remain compatible. New setups can select Qwen as agent/runner.
+- Qwen CLI uses Gemini-style arguments (`--approval-mode`, `--output-format stream-json`). `bare`, `reasoningEffort` and `nativeMultiAgent` selections are not supported (like Gemini).
+- Qwen routing profiles are configurable hypotheses, not authenticated benchmarks. Update installed packages and restart the dashboard/runner.
+
 ## 1.10.0 — 2026-09-06
 
 ### Added

package/README.md

@@ -1,15 +1,15 @@
 <div align="center">
 
-<h1><img src="https://raw.githubusercontent.com/HECer/yoke/v1.10.0/docs/assets/yoke-logo.png" alt="Yoke" width="100" height="63"></h1>
+<h1><img src="https://raw.githubusercontent.com/HECer/yoke/v1.12.0/docs/assets/yoke-logo.png" alt="Yoke" width="100" height="63"></h1>
 
-<!-- yoke:version:start -->1.10.0<!-- yoke:version:end -->
-<!-- yoke:tests:start -->1173<!-- yoke:tests:end -->
+<!-- yoke:version:start -->1.12.0<!-- yoke:version:end -->
+<!-- yoke:tests:start -->1213<!-- yoke:tests:end -->
 <!-- yoke:skills:start -->34<!-- yoke:skills:end -->
-<!-- yoke:agents:start -->Claude | Codex | Gemini<!-- yoke:agents:end -->
+<!-- yoke:agents:start -->Claude | Codex | Gemini | Qwen<!-- yoke:agents:end -->
 
-### One harness, three agents — and zero trust in "done."
+### One harness, four agents — and zero trust in "done."
 
-**Yoke** installs one curated canon of skills, **mechanical safety gates**, and tool wiring into any project — natively for **Claude Code, OpenAI Codex CLI, and Gemini CLI**. Its opt-in loop implements and verifies stories before committing. Independent review and browser proofs run when configured; screenshots and videos require the browser smoke gate.
+**Yoke** installs one curated canon of skills, **mechanical safety gates**, and tool wiring into any project — natively for **Claude Code, OpenAI Codex CLI, Gemini CLI, and Qwen Code**. Its opt-in loop implements and verifies stories before committing. Independent review and browser proofs run when configured; screenshots and videos require the browser smoke gate.
 
 [![npm](https://img.shields.io/npm/v/%40hecer%2Fyoke?logo=npm&color=CB3837)](https://www.npmjs.com/package/@hecer/yoke)
 [![npm downloads](https://img.shields.io/npm/dm/%40hecer%2Fyoke?logo=npm)](https://www.npmjs.com/package/@hecer/yoke)
@@ -17,8 +17,8 @@
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](#-license)
 ![Node](https://img.shields.io/badge/node-%E2%89%A520-339933?logo=node.js&logoColor=white)
 ![TypeScript](https://img.shields.io/badge/TypeScript-3178C6?logo=typescript&logoColor=white)
-![Tests](https://img.shields.io/badge/tests-1173%20defined-blue.svg)
-![Agents](https://img.shields.io/badge/agents-Claude%20%7C%20Codex%20%7C%20Gemini-8A2BE2)
+![Tests](https://img.shields.io/badge/tests-1213%20defined-blue.svg)
+![Agents](https://img.shields.io/badge/agents-Claude%20%7C%20Codex%20%7C%20Gemini%20%7C%20Qwen-8A2BE2)
 ![Built with TDD](https://img.shields.io/badge/built%20with-TDD%20%2B%20review-ff69b4.svg)
 
 **Install:** [`npm i -g @hecer/yoke`](https://www.npmjs.com/package/@hecer/yoke)
@@ -27,7 +27,7 @@
 
 > **TL;DR** — `yoke setup .` asks six questions and installs the native harness for your agent. `yoke new my-app --idea="..."` bootstraps a project and drafts its story backlog. `yoke loop run my-app --isolate --review` then implements it behind hard gates: **clean tree → acceptance criteria → your real tests green → an independent model approves → commit**. Add `--parallel=N` for dependency-aware workers, or declare a reference and add `--quality` for a bounded critic/repair gauntlet. If any blocking gate is red, nothing is committed. Proof lives in `.yoke/proof/<story>/`.
 
-**New in 1.10.0:** [dashboard search, filters and period comparisons](docs/DASHBOARD-EVOLUTION.md), [batch task assessments with separate planning models](docs/CAPABILITY-ROUTING.md), and [Windows sandbox preflight and process supervision](docs/WINDOWS-RUNNER-VALIDATION.md). Existing routing settings remain authoritative. See the [changelog](CHANGELOG.md) for migration and validation limits.
+**New in 1.12.0:** [Qwen Code hardening and explicit DeepSeek/Kimi API model profiles](docs/QWEN-MODEL-SUPPORT.md). Since 1.10.0, Yoke also includes [dashboard search, filters and period comparisons](docs/DASHBOARD-EVOLUTION.md), [batch task assessments with separate planning models](docs/CAPABILITY-ROUTING.md), and [Windows sandbox preflight and process supervision](docs/WINDOWS-RUNNER-VALIDATION.md). Existing routing settings remain authoritative. See the [changelog](CHANGELOG.md) for migration and validation limits.
 
 ### One dashboard, multiple projects
 
@@ -76,6 +76,19 @@ See [the 1.1 migration guide](docs/MIGRATING-TO-1.1.md) for setup/decision parit
 
 ---
 
+## Qwen Code, DeepSeek and Kimi
+
+Qwen Code supports Yoke's setup, routing, planning and review workflows. New Qwen
+setups use your configured model. To add DeepSeek and Kimi API model profiles:
+
+```sh
+yoke setup . --yes --agent=qwen --runner=qwen --model-provider=deepseek,kimi
+```
+
+Set `DEEPSEEK_API_KEY` and `MOONSHOT_API_KEY` in the environment. Presets include
+DeepSeek V4 Flash/Pro and Kimi K2.6/K2.7 Code/K3, executed through Qwen Code.
+See [setup, permissions, reasoning configuration and validation limits](docs/QWEN-MODEL-SUPPORT.md).
+
 ## Why Yoke exists
 
 Agentic coding in 2026 fails in four well-documented ways. Yoke answers each one **mechanically** — in code, not in a prompt the agent can ignore:
@@ -83,11 +96,11 @@ Agentic coding in 2026 fails in four well-documented ways. Yoke answers each one
 | The pain | What actually happens | What Yoke does about it |
 |---|---|---|
 | 🎭 **The verification gap** — *"agent says done, but it isn't"* | A success message can omit untested acceptance criteria. | The loop executes acceptance and project checks. Enabled review and browser gates must pass before a story lands. `yoke check` exposes unmapped outcomes as unverified. |
-| 🔀 **Three agents, three configs** | Teams hand-maintain `CLAUDE.md`, `AGENTS.md`, `GEMINI.md`, skills, and MCP wiring separately — copy-paste drift everywhere | **One canon → `yoke retrofit`** generates the idiomatic native artifacts for each agent. Change the canon once, re-retrofit everywhere. |
+| 🔀 **Four agents, four configs** | Teams hand-maintain `CLAUDE.md`, `AGENTS.md`, `GEMINI.md`, skills, and MCP wiring separately — copy-paste drift everywhere | **One canon → `yoke retrofit`** generates the idiomatic native artifacts for each agent. Change the canon once, re-retrofit everywhere. |
 | 🌀 **Overnight loops going off the rails** | Raw Ralph-loop users "wake up to broken codebases that don't compile" | Yoke is **"Ralph, but with gates"**: clean-worktree gate, acceptance-criteria gate, green-tests gate, review gate, per-story worktree isolation, idle-timeout watchdog, single-flight lock, commit integrity. |
 | 😵 **Review fatigue** | AI adoption nearly doubles PR volume and review time; humans start skimming | **`yoke review`**: a second model writes a schema-validated pass/fail verdict — chainable into verify, pre-push, or CI. Cross-model review catches what self-review misses. |
 
-**Who it's for:** anyone driving Claude Code, Codex CLI, or Gemini CLI on real projects — especially if you use more than one, want autonomous runs you can trust, or are tired of "done" meaning "probably". Greenfield (`yoke new`) and brownfield (`yoke retrofit`) both work.
+**Who it's for:** anyone driving Claude Code, Codex CLI, Gemini CLI, or Qwen Code on real projects — especially if you use more than one, want autonomous runs you can trust, or are tired of "done" meaning "probably". Greenfield (`yoke new`) and brownfield (`yoke retrofit`) both work.
 
 **Who it's not for:** if you want a chat pair-programmer with no process, you don't need a harness. Yoke is for shipping with discipline.
 
@@ -142,7 +155,7 @@ The canon is also packaged as a Claude Code plugin — the repo is its own marke
 /plugin install yoke@yoke
 ```
 
-That gives you all canon skills under the `yoke:` namespace (e.g. `yoke:tdd`, `yoke:review`) inside Claude Code — no retrofit needed. The `yoke` CLI (loop, gates, retrofit for Codex/Gemini) still comes from `npm i -g @hecer/yoke`. Gemini CLI users can likewise `gemini extensions install https://github.com/HECer/yoke`.
+That gives you all canon skills under the `yoke:` namespace (e.g. `yoke:tdd`, `yoke:review`) inside Claude Code — no retrofit needed. The `yoke` CLI (loop, gates, retrofit for Codex/Gemini/Qwen) still comes from `npm i -g @hecer/yoke`. Gemini CLI users can likewise `gemini extensions install https://github.com/HECer/yoke`.
 
 For Codex, no preinstalled skill is required: run `npx @hecer/yoke setup .` in a terminal, or
 ask Codex to run the six-question Yoke setup flow. The retrofit writes native skills to
@@ -169,7 +182,7 @@ Auto-upgrade is deliberately **not** the default: a gate harness shouldn't chang
 
 ## 🤖 Driving it through an agent
 
-Yoke is meant to be operated *by* your coding agent — after a retrofit, the agent has the skills, the safety policy, and the routing, so it knows the methodology. Copy-paste prompts (identical wording works for Claude Code, Codex CLI, and Gemini CLI):
+Yoke is meant to be operated *by* your coding agent — after a retrofit, the agent has the skills, the safety policy, and the routing, so it knows the methodology. Copy-paste prompts (identical wording works for Claude Code, Codex CLI, Gemini CLI, and Qwen Code):
 
 > **Set it up** — *"Set up Yoke in this project. Ask me the Yoke setup questions one at a time with your recommendation, then run `yoke setup . --yes` with the selected host, agents, code graph, loop, runner, and decision policy. Commit in my configured identity."*
 
@@ -193,10 +206,10 @@ Yoke's CLI is deterministic and chainable by design: an agent (or a shell `&&`)
 | `yoke projects add\|list\|remove` | Register a project, list registrations or remove a reference by ID | `0` · `2` invalid/unavailable |
 | `yoke check [dir] [--json] [--requirement=] [--protect [--refresh]]` | Execute acceptance checks or explicitly pin their infrastructure | `0` passed/pinned · `1` failed · `2` unverified/unavailable |
 | `yoke goal set\|run\|resume\|pause\|status\|handoff\|budget [dir]` | Durable objectives, provider handoff, protected checks and checkpoint budgets | run/resume: `0` complete · `1` unfinished · `2` unavailable |
-| `yoke setup [dir] [--yes] [--host=] [--agent=] [--runner=] [--code-graph=] [--decision-policy=] [--loop\|--no-loop] [--routing\|--no-routing]` | Shared six-question setup for Claude, Codex, and Gemini; routing defaults on for new setups and preserves explicit opt-outs | `0` · `1` invalid setup |
+| `yoke setup [dir] [--yes] [--host=] [--agent=] [--runner=] [--code-graph=] [--decision-policy=] [--loop\|--no-loop] [--routing\|--no-routing] [--model-provider=deepseek,kimi]` | Shared setup for Claude, Codex, Gemini and Qwen; optional DeepSeek/Kimi API profiles run through Qwen | `0` · `1` invalid setup |
 | `yoke validate [canonDir]` | Validate the canon (schema, frontmatter, templates) | `0` valid · `1` errors |
 | `yoke new <dir> [--idea=] [--agent=] [--runner=] [--loop]` | Greenfield bootstrap: git init → scaffold → retrofit → context → PRD (drafted from `--idea`) → committed | `0` · `1` usage / non-empty dir / draft failed (scaffold survives) · `2` draft agent unavailable |
-| `yoke retrofit [dir] [--agent=claude,codex,gemini\|all] [--code-graph=graphify\|serena] [--loop]` | Install/update the harness, non-destructively | `0` |
+| `yoke retrofit [dir] [--agent=claude,codex,gemini,qwen\|all] [--code-graph=graphify\|serena] [--loop]` | Install/update the harness for the selected agents, non-destructively | `0` |
 | `yoke prd draft [dir] --idea= [--runner=] [--force]` | Idea → 5–12 stories with testable acceptance criteria | `0` · `1` invalid/guarded · `2` agent unavailable |
 | `yoke prd check [dir]` | PRD lint gate (schema, dependencies, cycles, duplicate ids, acceptance) | `0` valid · `1` violations |
 | `yoke change add\|status [dir] [--idea=]` | Queue a change at any time; the loop turns it into append-only stories at the next safe boundary | `0` · `1` invalid inbox/request |
@@ -224,7 +237,7 @@ Three excellent projects, three different jobs. Honest version:
 | **Footprint** | Markdown skills (plugin) | ~230 MB with browser runtime; hourly auto-update | Node CLI + markdown canon; Playwright only if you use flow-smoke, resolved **from your project** |
 | **License** | MIT | MIT | MIT |
 
-**They compose — use all three where they're strongest.** Yoke's canon *ships* the superpowers methodology natively for all three agents (13 skills, [attributed](canon/skills/ATTRIBUTION.md)). And if gstack is installed, `yoke retrofit` detects it and adds a routing note to `CLAUDE.md` telling Claude to prefer gstack's live-browser `/qa`, `/cso`, and ship pipeline for what Yoke deliberately doesn't bundle — no dependency, no conflict, and Codex/Gemini artifacts stay uniform.
+**They compose — use all three where they're strongest.** Yoke's canon *ships* the superpowers methodology natively for all four agents (13 skills, [attributed](canon/skills/ATTRIBUTION.md)). And if gstack is installed, `yoke retrofit` detects it and adds a routing note to `CLAUDE.md` telling Claude to prefer gstack's live-browser `/qa`, `/cso`, and ship pipeline for what Yoke deliberately doesn't bundle — no dependency, no conflict, and Codex/Gemini artifacts stay uniform.
 
 **Choose Yoke when** you run more than one agent, want autonomy you can audit (gates + proofs + logs), or want one place to maintain your team's methodology. **Choose gstack when** you live 100% in Claude Code and want the deepest interactive browser QA. **Choose superpowers when** you want the methodology alone, interactively, in Claude Code — or just use it *through* Yoke.
 
@@ -240,10 +253,12 @@ flowchart TD
     Skill --> Claude["Claude Code<br/>.claude/skills · .mcp.json · hook"]
     Skill --> Codex["Codex CLI<br/>AGENTS.md · config.toml · RTK.md"]
     Skill --> Gemini["Gemini CLI<br/>GEMINI.md · commands · settings.json"]
+    Skill --> Qwen["Qwen Code<br/>QWEN.md · skills · settings.json"]
     Loop["🤖 yoke loop — autonomous Ralph loop<br/>gates · verify · review · isolation · proofs"]
     Claude -. drives .-> Loop
     Codex -. drives .-> Loop
     Gemini -. drives .-> Loop
+    Qwen -. drives .-> Loop
 ```
 
 Three layers — **Canon** (`yoke validate`) → **Retrofit** (`yoke retrofit`) → **Loop** (`yoke loop`) — on top of a durable **Context layer** (`yoke context`).
@@ -255,6 +270,7 @@ Three layers — **Canon** (`yoke validate`) → **Retrofit** (`yoke retrofit`)
 | **Claude** | Complete skill packages under `.claude/skills/` (including referenced resources), `AGENTS.md`, `CLAUDE.md`, `.mcp.json` (code-graph + Playwright), and an rtk `PreToolUse` hook when WSL is available |
 | **Codex** | Complete skill packages under `.agents/skills/`, per-skill implicit-invocation policy, `AGENTS.md`, `RTK.md`, `.codex/config.toml`, native hooks, reusable `.codex/agents/*.toml`, and package plugin metadata |
 | **Gemini** | Complete skill packages under `.gemini/skills/`, an auto-invocation index, `GEMINI.md`, `.gemini/commands/*.toml`, and `.gemini/settings.json` (MCP + `AGENTS.md` context) |
+| **Qwen** | Complete skill packages under `.qwen/skills/`, `QWEN.md`, `.qwen/settings.json`, native invocation restrictions, and an RTK PreToolUse retry guard |
 
 > **rtk integration:** Claude receives its PreToolUse hook; Codex receives a native hook adapter around `rtk hook check`; Gemini retains instruction-mode fallback where its CLI has no equivalent command-rewrite lifecycle.
 
@@ -271,10 +287,10 @@ Three layers — **Canon** (`yoke validate`) → **Retrofit** (`yoke retrofit`)
 
 `yoke retrofit` installs all of these into each agent natively. Provenance is credited in [`canon/skills/ATTRIBUTION.md`](canon/skills/ATTRIBUTION.md).
 
-To stop overlapping skills from auto-invoking against each other, `canon/AGENTS.md` carries a **skill routing & precedence** block (methodology before role; one canonical entrypoint per concern — e.g. pre-merge code review is always `review`), emitted into all three agents.
+To stop overlapping skills from auto-invoking against each other, `canon/AGENTS.md` carries a **skill routing & precedence** block (methodology before role; one canonical entrypoint per concern — e.g. pre-merge code review is always `review`), emitted into all four agents.
 
 Each manifest entry also declares `invocation: auto|manual`. Retrofit translates that intent into
-the provider's native controls: Claude disables model invocation for manual skills, Codex writes
+the provider's native controls: Claude and Qwen disable model invocation for manual skills, Codex writes
 `agents/openai.yaml`, and Gemini lists only automatic skills in its generated index. Validation
 rejects conflicting package metadata and broken local Markdown links before anything is installed.
 
@@ -573,13 +589,13 @@ parent remains the strong planner/controller. Before each bounded story it recei
 story, acceptance criteria, and at most three eligible worker profiles, then returns one
 machine-readable choice. The worker can be a cheaper/faster Claude, Codex, or Gemini profile;
 `SELF` keeps difficult work on the parent. Explicit project rules skip the controller.
-Loop runners disable native delegation in Codex, Claude and Gemini so it cannot multiply
+Loop runners disable native delegation in Codex, Claude, Gemini and Qwen so it cannot multiply
 the Yoke worker budget. Integration retains its execution slot until the candidate lands.
 
 **Provider support:** adaptive routing uses Yoke's shared provider adapter and works with Claude
-Code, Codex CLI, and Gemini CLI, including mixed-provider worker lists. Internal contract tests
-cover invocation and routing behavior for all three providers. The measured performance evidence
-below is intentionally **Codex-only**; it does not claim equivalent Claude or Gemini savings
+Code, Codex CLI, Gemini CLI, and Qwen Code, including mixed-provider worker lists. Internal contract tests
+cover invocation and routing behavior for all four providers. The measured performance evidence
+below is intentionally **Codex-only**; it does not claim equivalent Claude, Gemini or Qwen savings
 until authenticated, repeated in-the-wild runs exist for those providers.
 
 ```yaml
@@ -673,7 +689,7 @@ for example:
 An agent can read that ordinary file when the preview is insufficient; nothing is injected into
 later stories automatically. Repeated identical failures reuse the same content-addressed path.
 Successful gate output is discarded as before. This affects only commands executed by Yoke's own
-gates. It does **not** intercept tool output generated internally by Claude Code, Codex, or Gemini,
+gates. It does **not** intercept tool output generated internally by Claude Code, Codex, Gemini, or Qwen,
 so benchmark ratios for this feature are not provider-token or billing claims.
 
 Command capture is capped at 16 MiB per stdout/stderr stream. Exceeding that quota fails the gate
@@ -897,7 +913,7 @@ release provenance.
 ## 🧪 Development
 
 ```bash
-npm test          # vitest (1173 tests)
+npm test          # vitest (1213 tests)
 npm run build     # tsc, no emit errors
 npm run yoke -- validate canon
 ```

package/canon/manifest.yaml

@@ -1,6 +1,6 @@
 name: yoke-canon
-version: 1.7.0
-agents: [claude, codex, gemini]
+version: 1.12.0
+agents: [claude, codex, gemini, qwen]
 skills:
   - { id: tdd, path: skills/tdd, kind: methodology, invocation: auto }
   - { id: yoke-retrofit, path: skills/yoke-retrofit, kind: methodology, invocation: auto }

package/canon/tools/qwen-rtk-hook.mjs

@@ -0,0 +1,25 @@
+#!/usr/bin/env node
+// PreToolUse guard requesting a retry through RTK. Never evaluates or executes tool commands.
+import { readFileSync } from 'node:fs'
+
+const record = value => value !== null && typeof value === 'object' && !Array.isArray(value)
+try {
+  const event = JSON.parse(readFileSync(0, 'utf8'))
+  if (!record(event) || typeof event.tool_name !== 'string' ||
+      (event.hook_event_name !== undefined && event.hook_event_name !== 'PreToolUse')) throw new Error('event')
+  let response = {}
+  if (event.tool_name === 'run_shell_command') {
+    if (!record(event.tool_input) || typeof event.tool_input.command !== 'string' || !event.tool_input.command.trim() || event.tool_input.command.includes('\0')) throw new Error('command')
+    const command = event.tool_input.command
+    // Restrict rewriting to simple supported invocations. Leave shell syntax,
+    // quoted executables, assignments and existing RTK wrappers untouched.
+    if (!/[;&|<>`$\r\n()]/u.test(command) && /^\s*(?:git|rg|npm|npx|cargo|pytest|go|docker|kubectl)\s/u.test(command)) {
+      const rewritten = command.trimStart().replace(/^rg\s/u, 'grep ')
+      response = { hookSpecificOutput: { hookEventName: 'PreToolUse', permissionDecision: 'deny', permissionDecisionReason: `Retry this command through RTK: rtk ${rewritten}` } }
+    }
+  }
+  process.stdout.write(JSON.stringify(response) + '\n')
+} catch {
+  process.stderr.write('Invalid Qwen PreToolUse hook input\n')
+  process.exitCode = 2
+}

package/dist/agents/contracts.js

@@ -1,5 +1,5 @@
 import { z } from 'zod';
-export const AgentSchema = z.enum(['claude', 'codex', 'gemini']);
+export const AgentSchema = z.enum(['claude', 'codex', 'gemini', 'qwen']);
 export const PermissionProfileSchema = z.enum(['safe', 'unsafe', 'read-only']);
 export const ModelSelectionSchema = z.object({
     model: z.string().regex(/^[A-Za-z0-9][A-Za-z0-9._:/-]{0,127}$/).optional(),

package/dist/agents/host.js

@@ -7,12 +7,16 @@ export function detectHostAgent(env = process.env) {
         return 'claude';
     if (env.GEMINI_CLI)
         return 'gemini';
+    if (env.QWEN_CODE || env.QWEN_CODE_SESSION_ID || env.QWEN_CLI)
+        return 'qwen';
     if (env.CODEX_HOME)
         return 'codex';
     if (env.CLAUDE_CONFIG_DIR)
         return 'claude';
     if (env.GEMINI_CLI_HOME)
         return 'gemini';
+    if (env.QWEN_CLI_HOME)
+        return 'qwen';
     return undefined;
 }
 export function resolveRunnerAgent(config, explicit, host) {

package/dist/agents/providers.js

@@ -17,6 +17,13 @@ const argsFor = (agent, permissions) => {
         // Automatic review already selects workspace-write and conflicts with --sandbox.
         return ['exec', '--approve-for-me', '--json'];
     }
+    // Qwen Code uses its own approval modes and native tool exclusions.
+    if (agent === 'qwen') {
+        if (permissions === 'unsafe')
+            return ['--yolo', '--output-format', 'stream-json'];
+        const approval = permissions === 'read-only' ? 'plan' : 'auto-edit';
+        return ['--approval-mode', approval, '--sandbox', '--output-format', 'stream-json', ...(permissions === 'safe' ? ['--allowed-tools', 'run_shell_command'] : [])];
+    }
     if (permissions === 'unsafe')
         return ['--yolo', '--output-format', 'stream-json'];
     const approval = permissions === 'read-only' ? 'plan' : 'auto_edit';
@@ -30,6 +37,12 @@ export function buildProviderInvocation(agent, prompt, cwd, permissions = 'safe'
         throw new Error('Gemini does not support the reasoningEffort selection');
     if (agent === 'gemini' && parsedSelection.nativeMultiAgent === true)
         throw new Error('Gemini does not support enabling the nativeMultiAgent selection');
+    if (agent === 'qwen' && parsedSelection.bare)
+        throw new Error('Qwen does not support the bare startup selection');
+    if (agent === 'qwen' && parsedSelection.reasoningEffort)
+        throw new Error('Qwen does not support the reasoningEffort selection');
+    if (agent === 'qwen' && parsedSelection.nativeMultiAgent === true)
+        throw new Error('Qwen does not support enabling the nativeMultiAgent selection');
     const args = argsFor(agent, permissions);
     if (output.schemaFile !== undefined || output.jsonSchema !== undefined) {
         if (agent === 'codex' && output.schemaFile && output.jsonSchema === undefined) {
@@ -46,8 +59,18 @@ export function buildProviderInvocation(agent, prompt, cwd, permissions = 'safe'
         else
             throw new Error(`${agent} structured output schema requires ${agent === 'codex' ? 'schemaFile' : agent === 'claude' ? 'jsonSchema' : 'a supported native schema option (unavailable)'}`);
     }
-    if (parsedSelection.model)
-        args.push('--model', parsedSelection.model);
+    if (parsedSelection.model) {
+        const qualified = agent === 'qwen' && parsedSelection.model.includes('::')
+            ? parsedSelection.model.match(/^(openai|anthropic|gemini|vertex-ai|qwen-oauth)::(.+)$/u) : undefined;
+        if (agent === 'qwen' && parsedSelection.model.includes('::') && !qualified)
+            throw Error('Invalid Qwen auth/model selector');
+        if (qualified) {
+            const model = ModelSelectionSchema.shape.model.parse(qualified[2]);
+            args.push('--auth-type', qualified[1], '--model', model);
+        }
+        else
+            args.push('--model', parsedSelection.model);
+    }
     if (parsedSelection.reasoningEffort) {
         if (agent === 'claude')
             args.push('--effort', parsedSelection.reasoningEffort);
@@ -58,6 +81,9 @@ export function buildProviderInvocation(agent, prompt, cwd, permissions = 'safe'
         args.push('--disable', 'multi_agent');
     if (agent === 'claude' && parsedSelection.nativeMultiAgent === false)
         args.push('--disallowedTools', 'Agent', 'Task', 'TeamCreate', 'SendMessage');
+    if (agent === 'qwen' && parsedSelection.nativeMultiAgent === false) {
+        args.push('--exclude-tools', 'agent', 'task', 'create_sub_session', 'team_create', 'send_message');
+    }
     if (parsedSelection.bare) {
         if (agent === 'codex')
             args.push('--ignore-user-config');

package/dist/agents/telemetry.js

@@ -22,6 +22,8 @@ export function parseProviderResult(agent, output) {
         if (direct !== undefined)
             return direct;
     }
+    if (agent === 'qwen')
+        return parseQwenResult(output);
     const fragments = [];
     for (const line of output.split(/\r?\n/u)) {
         const parsed = parseJson(line);
@@ -63,6 +65,34 @@ export function parseProviderResult(agent, output) {
     }
     return null;
 }
+/** Qwen uses assistant content blocks and result envelopes, not Gemini messages. */
+function parseQwenResult(output) {
+    let candidate = null;
+    for (const line of output.split(/\r?\n/u)) {
+        const parsed = parseJson(line);
+        if (!parsed.ok || !isRecord(parsed.value))
+            continue;
+        const event = parsed.value;
+        if (event.parent_tool_use_id != null)
+            continue;
+        if (event.type === 'result') {
+            if (event.is_error === true) {
+                candidate = null;
+                continue;
+            }
+            const structured = directMachineResult(event.structured_result);
+            const result = typeof event.result === 'string' ? parseJson(event.result) : { ok: false };
+            candidate = structured ?? (result.ok ? directMachineResult(result.value) : undefined) ?? null;
+        }
+        else if (event.type === 'assistant' && isRecord(event.message) && Array.isArray(event.message.content)) {
+            const text = event.message.content.filter(isRecord).filter(part => part.type === 'text' && typeof part.text === 'string').map(part => part.text).join('');
+            const result = parseJson(text);
+            if (result.ok)
+                candidate = directMachineResult(result.value) ?? candidate;
+        }
+    }
+    return candidate;
+}
 export function parseProviderTelemetry(agent, lines) {
     let inputTokens;
     let cachedInputTokens;
@@ -83,6 +113,8 @@ export function parseProviderTelemetry(agent, lines) {
         if (!parsed || typeof parsed !== 'object' || Array.isArray(parsed))
             continue;
         const event = parsed;
+        if (agent === 'qwen' && event.parent_tool_use_id != null)
+            continue;
         const message = event.message && typeof event.message === 'object' ? event.message : undefined;
         const stats = event.stats && typeof event.stats === 'object' ? event.stats : undefined;
         const usage = (event.usage && typeof event.usage === 'object'
@@ -101,12 +133,12 @@ export function parseProviderTelemetry(agent, lines) {
         // Older JSON stats only provide model-local token objects. Sum a field
         // only when every model measured it; a missing measurement is not zero.
         let source = usage ?? nestedModelTokens ?? modelUsage;
-        if (agent === 'gemini' && modelEntries.length > 0) {
+        if ((agent === 'gemini' || agent === 'qwen') && modelEntries.length > 0) {
             reportedModels = modelEntries.map(([name]) => name);
             model = firstModel?.[0];
             const fields = {
-                input_tokens: ['input_tokens', 'inputTokens', 'promptTokenCount', 'input'],
-                output_tokens: ['output_tokens', 'outputTokens', 'candidatesTokenCount', 'output'],
+                input_tokens: ['input_tokens', 'inputTokens', 'promptTokenCount', 'input', 'prompt'],
+                output_tokens: ['output_tokens', 'outputTokens', 'candidatesTokenCount', 'output', 'candidates'],
                 cached_input_tokens: ['cached_input_tokens', 'cachedInputTokens', 'cachedContentTokenCount', 'cached'],
                 reasoning_output_tokens: ['reasoning_output_tokens', 'reasoningOutputTokens', 'thoughtsTokenCount', 'thoughts'],
             };

package/dist/canon/manifest.js

@@ -1,7 +1,7 @@
 import { z } from 'zod';
 import { parse } from 'yaml';
 import { readFileSync } from 'node:fs';
-export const AgentSchema = z.enum(['claude', 'codex', 'gemini']);
+export const AgentSchema = z.enum(['claude', 'codex', 'gemini', 'qwen']);
 export const InvocationSchema = z.enum(['auto', 'manual']);
 export const SkillEntrySchema = z.object({
     id: z.string().min(1),