@martinffx/atelier 3.6.0 → 3.7.0

package/CHANGELOG.md

@@ -4,6 +4,15 @@
 
 - **breaking:** rename the `recon` agent to `sentinel`. Existing Atelier configuration is migrated automatically and legacy generated files are removed during the next command.
 
+## [3.7.0](https://github.com/martinffx/atelier/compare/v3.6.0...v3.7.0) (2026-11-07)
+
+
+### Features
+
+* **agents:** make keymaker the implementer of settled designs ([6c7fb33](https://github.com/martinffx/atelier/commit/6c7fb33463006d64f68f6bb2d7056347b3eb7bbf))
+* **claude:** set a 190k auto-compact window under opusplan ([a9f1ed1](https://github.com/martinffx/atelier/commit/a9f1ed156f10fbde547dcffe9b31b33409a204f3))
+* **opencode:** refresh model catalogs and add openrouter and google providers ([#90](https://github.com/martinffx/atelier/issues/90)) ([85ed86b](https://github.com/martinffx/atelier/commit/85ed86b5a070f08624c8f4dab2445ebeb8b73db3))
+
 ## [3.6.0](https://github.com/martinffx/atelier/compare/v3.5.0...v3.6.0) (2026-10-25)
 
 

package/README.md

@@ -114,24 +114,26 @@ The CLI also configures four specialist agents for Claude Code, OpenCode, Codex,
 | **Sentinel** | Read-only codebase reconnaissance and review triage |
 | **Oracle** | Requirements, trade-offs, and critical analysis of changes |
 | **Architect** | Domain modelling, system design, and architecture review |
-| **Keymaker** | Fast scaffolding of files and boilerplate from templates and existing patterns |
+| **Keymaker** | Implements settled designs: code, business logic and tests, so the main conversation stays small |
 
-Each subagent launch explicitly selects one of these four by matching the task to its description; the harness default is never used. Keymaker is the only agent with write access in Codex, and it uses Sentinel's model and thinking defaults in every configuration.
+Each subagent launch explicitly selects one of these four by matching the task to its description; the harness default is never used. Keymaker is the only agent with write access in Codex. It runs on each harness's build-tier model (Sonnet on Claude Code), not the cheap Sentinel model.
 
 Run `npx @martinffx/atelier@latest --help` for CLI commands and options. Each skill contains its own operating instructions and loads when its context applies.
 
 ## Models and thinking
 
-New configurations use the following defaults (September 2026). Each entry is **model / thinking**.
+New configurations use the following defaults (November 2026). Each entry is **model / thinking**.
 
 | Configuration | Sentinel | Oracle | Architect | Plan | Build |
 |---|---|---|---|---|---|
 | Claude Code | Haiku 5.5 / default | Opus 5.5 / xhigh | Opus 5.5 / high | Opus 5.5 / high | Sonnet 5 / high |
 | Codex | Luna 6 / low | Astra 6 / high | Astra 6 / xhigh | Sol 6 / xhigh | Sol 6 / high |
 | OpenCode / OpenAI | Luna 6 / low | Astra 6 / high | Astra 6 / xhigh | Astra 6 / xhigh | Sol 6 / high |
-| OpenCode / Bedrock | Haiku 4.5 / default | Fable 5.1 / high | Opus 5.5 / high | Opus 5.5 / high | Sonnet 5 / high |
+| OpenCode / Bedrock | Haiku 5.5 / low | Fable 5.1 / high | Opus 5.5 / high | Opus 5.5 / high | Sonnet 5 / high |
 | OpenCode / Zen | GLM 5.3 Flash / low | Kimi K3 / high | GLM 5.3 / high | GLM 5.3 / high | DeepSeek V4.1 Flash / high |
 | OpenCode / Go | GLM 5.3 Flash / low | MiMo V2.6 Pro / on | GLM 5.3 / high | GLM 5.3 / high | DeepSeek V4.1 Flash / high |
+| OpenCode / OpenRouter | GLM 5.3 Flash / low | Kimi K3.1 Pro / high | GLM 5.3 / high | GLM 5.3 / high | DeepSeek V4.1 Flash (1017) / high |
+| OpenCode / Google | Gemini 3.5 Flash Lite / low | Gemini 4 Argon / high | Gemini 4 Argon / high | Gemini 4 Argon / high | Gemini 3.8 Flash / high |
 
 Sentinel favors inexpensive reconnaissance. Oracle needs requirements synthesis and judgment; Architect and Plan need technical depth and completeness. Build balances capability and execution cost. These are starting configurations, informed by [Artificial Analysis](https://artificialanalysis.ai/models), rather than measured winners for Atelier's specific roles. Benchmark effort levels and provider speeds may differ from these defaults.
 
@@ -141,9 +143,9 @@ Run `atelier update --harness <claude|codex|opencode>` to choose models and thin
 
 Thinking is stored as `agents[].thinking`, `build_thinking`, and `plan_thinking` in `~/.atelier/config.json`. Values are model-specific: `default`, `off`, `on`, `low`, `medium`, `high`, `xhigh`, or `max`. The picker offers only verified capabilities. Custom models and catalog models whose controls have not been verified offer `default` only. Missing fields preserve legacy behavior; **Keep existing behavior** preserves that omission when editing. In particular, old Codex configurations retain medium agent/Build effort and high Plan effort. An explicit `default` omits Atelier's native effort override; other harness settings can still apply.
 
-- **Claude Code:** `opusplan` switches Opus/Sonnet by mode. Other session models share one thinking setting across modes. Atelier writes canonical `modelSettings.<model>.effortLevel` entries and subagent `effort`. Session `max` cannot be persisted, though supported subagents can use it. Use Claude Code **2.1.280 or later** for Opus 5.5 and the current aliases. Atelier also sets `disableClaudeAiConnectors: true`, `syncClaudeAiSkills: false` and `syncClaudeAiPlugins: false` in `~/.claude/settings.json`, so nothing from claude.ai is synced into Claude Code; `atelier remove` clears them unless you changed them. It also disables the Claude Code built-ins that clash with Atelier skills (`disableBundledSkills`, `enableArtifact: false`, `disableWorkflows`, deny rules for `Artifact`, `ArtifactComments`, `ArtifactData`, `Workflow` and `DesignSync`, and `skillOverrides` turning off the `/security-review` and `/ultrareview` commands). Interactive `init` also offers a checklist of optional built-ins to disable, none selected by default: `cron` (`/loop` and the Cron tools), `chrome` (the claude-in-chrome MCP server) and `remote-trigger` (`RemoteTrigger`). `atelier remove` and unticking an option undo these the same way. [Claude configuration](https://code.claude.com/docs/en/model-config)
+- **Claude Code:** `opusplan` switches Opus/Sonnet by mode. Other session models share one thinking setting across modes. Atelier writes canonical `modelSettings.<model>.effortLevel` entries and subagent `effort`. Session `max` cannot be persisted, though supported subagents can use it. Under `opusplan`, Atelier also writes `modelSettings.<model>.autoCompactWindow: 190000` for both models: Claude Code silently plans on Sonnet once a turn passes 200k tokens, but only auto-compacts near 967k ([anthropics/claude-code#95206](https://github.com/anthropics/claude-code/issues/95206)). A window you set yourself is kept, and `atelier remove` clears only the 190000 value. Use Claude Code **2.1.280 or later** for Opus 5.5 and the current aliases. Atelier also sets `disableClaudeAiConnectors: true`, `syncClaudeAiSkills: false` and `syncClaudeAiPlugins: false` in `~/.claude/settings.json`, so nothing from claude.ai is synced into Claude Code; `atelier remove` clears them unless you changed them. It also disables the Claude Code built-ins that clash with Atelier skills (`disableBundledSkills`, `enableArtifact: false`, `disableWorkflows`, deny rules for `Artifact`, `ArtifactComments`, `ArtifactData`, `Workflow` and `DesignSync`, and `skillOverrides` turning off the `/security-review` and `/ultrareview` commands). Interactive `init` also offers a checklist of optional built-ins to disable, none selected by default: `cron` (`/loop` and the Cron tools), `chrome` (the claude-in-chrome MCP server) and `remote-trigger` (`RemoteTrigger`). `atelier remove` and unticking an option undo these the same way. [Claude configuration](https://code.claude.com/docs/en/model-config)
 - **Codex:** Plan and Build share the selected session model, with separate reasoning settings. Oracle and Architect use independent Astra configurations. `off` maps to native `none`, available on Sol/Luna but not Astra. [Codex configuration](https://developers.openai.com/codex/config-reference/)
-- **OpenCode:** all five assignments are independent. Use **1.18.30 or later** for GPT-6 integration. Atelier sends explicit provider options rather than relying on automatically discovered reasoning variants: `reasoningEffort` for OpenAI-compatible routes, `effort` for Anthropic, and `reasoningConfig` for Bedrock. GLM 5.3 supports low/high/max; it does not support medium. Bedrock defaults use US inference profiles with `us-east-1`. [OpenCode agents](https://opencode.ai/docs/agents/)
+- **OpenCode:** all five assignments are independent. Use **1.18.30 or later** for GPT-6 integration. Atelier sends explicit provider options rather than relying on automatically discovered reasoning variants: `reasoningEffort` for OpenAI-compatible routes, `effort` for Anthropic, and `reasoningConfig` for Bedrock, `reasoning.effort` for OpenRouter and `thinkingConfig.thinkingLevel` for Google. GLM 5.3 supports low/high/max; it does not support medium. Bedrock defaults use US inference profiles with `us-east-1`. Go's Muse Spark 1.3 Contributor only works if the workspace allows paid endpoints that train on request data. [OpenCode agents](https://opencode.ai/docs/agents/)
 - **MiMo:** `on`/`off` maps to `thinking.type: enabled`/`disabled`. There are no graded effort tiers. OpenCode agent files omit the old fixed temperature override so provider defaults apply. [MiMo API](https://mimo.mi.com/docs/en-US/api/chat/openai-api)
 
 Atelier does not upgrade installed harness CLIs. Request-option serialization was checked against OpenCode's provider SDK versions with intercepted requests; this verifies parameter encoding, not upstream availability, quota, or role quality.

package/agents/keymaker.md

@@ -1,36 +1,44 @@
 ---
 name: keymaker
-description: Fast scaffolder. Use proactively to create files, directories and boilerplate from templates or existing project patterns, or to make simple pattern-following edits; reports exactly what changed. Does not design, review or implement business logic.
+description: Implementer. MUST BE USED for every code change once the design or plan is settled, including new files, features, business logic, refactors, test writing, bug fixes with a known cause, and multi-file edits. Delegate implementation to Keymaker instead of editing in the main conversation, so the main conversation's context stays small. Give it the design excerpt, owned files and validation command. It writes tests first, implements, runs validation and reports files changed and test results. It makes no design decisions; it stops and asks.
 ---
 
-You are the **Keymaker**, a fast scaffolding agent. Your job is to create the files and directories others need, by copying existing project patterns or applying templates exactly. You work quickly, follow the pattern you are given, and never improvise design.
+You are the **Keymaker**, the implementer. The design is settled and someone has handed you one task to build. You write the code and tests, run the validation, and report back, so the main conversation never has to carry the implementation.
+
+Delegating to you is the default path for writing code. The main conversation edits directly only for trivial one-line changes and stateful debugging.
 
 ## Role
 
-- Create files and directory structures from templates or existing project patterns
-- Apply variable substitutions consistently across everything you create
-- Make simple, mechanical, pattern-following edits (renames, boilerplate, wiring)
-- Report exactly what was created or changed
+- Implement one task from the design excerpt you are given: files, business logic, wiring, and tests
+- Follow existing project patterns and conventions rather than inventing new ones
+- Write the test first when behaviour changes (stub → test → fix), then make it pass
+- Run the task's validation command and fix what fails, within the files you own
+- Report exactly what changed, and what the validation showed
 
 ## Skills
 
-Before beginning work, scan your environment for relevant skills and load any that apply to the task at hand. Use them to follow the project's conventions for structure and naming.
+Before beginning work, scan your environment for relevant skills and load any that apply to the task at hand. Use them to follow the project's conventions for structure, naming and testing.
+
+## Inputs
+
+You should be given: the task and its requirements, the design excerpt it implements, the files you own, and the validation command. If any of these is missing or ambiguous, stop and ask. Do not guess your way around a gap in the design.
 
 ## Checklist
 
 Before finishing, confirm you have:
 
-- [ ] Verified each target file does not already exist before creating it
-- [ ] Created parent directories as needed
-- [ ] Applied all relevant variable substitutions when using templates
-- [ ] Matched the pattern you were pointed at (naming, structure, imports)
-- [ ] Reported every file created or modified
+- [ ] Stayed inside the files you were told you own
+- [ ] Matched the pattern you were pointed at (naming, structure, imports, test style)
+- [ ] Written or updated tests for any changed behaviour
+- [ ] Run the validation command and read its output
+- [ ] Reported every file created or modified, plus the validation result (pass or the failing output)
 
 ## Boundaries
 
-- DO create files and directories as requested
-- DO follow the pattern you are given rather than inventing one
+- DO implement the task, including business logic, within your owned files
+- DO write and update tests
+- DO stop and ask when the design is unclear, contradicts the code, or a change would reach outside your owned files
 - DON'T make design or architectural decisions (that's `architect`)
-- DON'T implement business logic or review code
-- DON'T conduct requirements discovery (that's `oracle`)
-- NEVER overwrite existing files unless explicitly told to
+- DON'T conduct requirements discovery or challenge scope (that's `oracle`)
+- DON'T review code beyond checking your own work
+- NEVER overwrite files outside your owned list, and never silently work around a failing test by weakening it

package/dist/atelier.js

@@ -32104,17 +32104,33 @@ function resolveBasePath(harness) {
 var THINKING_LEVELS = ["default", "off", "on", "low", "medium", "high", "xhigh", "max"];
 function thinkingChoices(harness, model, session = false) {
   const id = model.split("/").at(-1).toLowerCase();
-  if (/gpt-6-(astra|sol|luna)$/.test(id)) {
-    return ["default", ...id.endsWith("astra") ? [] : ["off"], "low", "medium", "high", "xhigh", "max"];
+  if (/gpt-6(\.1)?-(astra|sol|luna|terra)$/.test(id)) {
+    return ["default", .../astra$|6\.1-sol$/.test(id) ? [] : ["off"], "low", "medium", "high", "xhigh", "max"];
   }
-  if (/^(opus|opusplan|sonnet|haiku|fable|best)$/.test(id) || /claude-(opus-5-5|fable-5-1|sonnet-5-5|haiku-5-5)$/.test(id)) {
+  if (/^(opus|opusplan|sonnet|haiku|fable|best)$/.test(id) || /claude-(opus-5[-.]5|fable-5[-.]1|sonnet-5[-.]5|haiku-5[-.]5)$/.test(id)) {
     return ["default", "low", "medium", "high", "xhigh", ...harness === "claude" && session ? [] : ["max"]];
   }
   if (harness === "opencode") {
-    if (/^(glm-5\.3(-flash)?|kimi-k3|deepseek-v4(\.1-flash|-pro))$/.test(id))
+    if (/^(glm-5\.3(-flash)?|kimi-k3(\.1-pro)?|deepseek-v4(\.1-flash(-1017)?|-pro))$/.test(id))
       return ["default", "low", "high", "max"];
     if (/^(mimo-v2\.6-(pro|flash)(-free)?|minimax-m3(-free)?)$/.test(id))
       return ["default", "off", "on"];
+    if (id === "minimax-m3.1-flash")
+      return ["default", "low", "medium", "high", "xhigh", "max"];
+    if (/^(gemini-4-argon(-contributor)?|gemini-3\.[5-8]-flash(-lite)?)$/.test(id))
+      return ["default", "low", "medium", "high"];
+    if (id === "grok-4.7")
+      return ["default", "low", "medium", "high", "xhigh"];
+    if (/^mistral-large-4(-0)?$/.test(id))
+      return ["default", "off", "high"];
+    if (/^inkling(-2)?(-small)?(-free)?$/.test(id))
+      return ["default", "off", "low", "medium", "high", "max"];
+    if (/^muse-(spark-1\.3(-contributor)?|wonder(-1\.0)?)$/.test(id))
+      return ["default", "low", "medium", "high", "xhigh", "max"];
+    if (id === "muse-spark-open")
+      return ["default", "low", "medium", "high", "xhigh"];
+    if (/^nemotron-3\.5-super-vl(-free)?$/.test(id))
+      return ["default", "off", "low", "medium", "high", "xhigh"];
   }
   return ["default"];
 }
@@ -32136,7 +32152,7 @@ function claudeModel(model) {
 function claudeModeModels(model) {
   return model === "opusplan" ? { build: claudeModel("sonnet"), plan: claudeModel("opus") } : { build: claudeModel(model), plan: claudeModel(model) };
 }
-var OPENCODE_THINKING_KEYS = ["reasoningEffort", "thinking", "effort", "reasoningConfig"];
+var OPENCODE_THINKING_KEYS = ["reasoningEffort", "thinking", "effort", "reasoningConfig", "reasoning", "thinkingConfig"];
 function openCodeThinkingOptions(model, thinking) {
   assertThinking("opencode", model, thinking);
   if (!thinking || thinking === "default")
@@ -32144,10 +32160,17 @@ function openCodeThinkingOptions(model, thinking) {
   if (model.startsWith("amazon-bedrock/")) {
     return { reasoningConfig: { type: "adaptive", maxReasoningEffort: thinking, display: "summarized" } };
   }
+  if (model.startsWith("openrouter/")) {
+    if (thinking === "on" || thinking === "off")
+      return { reasoning: { enabled: thinking === "on" } };
+    return { reasoning: { effort: thinking } };
+  }
+  if (model.startsWith("google/"))
+    return { thinkingConfig: { thinkingLevel: thinking } };
   if (model.includes("claude-"))
     return { effort: thinking };
   if (thinking === "on" || thinking === "off") {
-    if (model.includes("gpt-6-"))
+    if (/gpt-6|mistral-large|inkling|nemotron/.test(model))
       return { reasoningEffort: "none" };
     return { thinking: { type: thinking === "off" ? "disabled" : model.includes("minimax") ? "adaptive" : "enabled" } };
   }
@@ -32293,7 +32316,7 @@ function formatFileList(files) {
 `);
 }
 function guardProvider(provider) {
-  if (!provider || !["opencode-zen", "opencode-go", "amazon-bedrock", "openai"].includes(provider)) {
+  if (!provider || !["opencode-zen", "opencode-go", "amazon-bedrock", "openai", "openrouter", "google"].includes(provider)) {
     throw new InvalidConfigError("A valid provider must be selected for OpenCode");
   }
   return provider;
@@ -32561,7 +32584,7 @@ var SimpleConfigSchema = objectType({
 var CursorConfigSchema = objectType({
   agents: arrayType(AgentSchema).min(1)
 }).strict();
-var OPENCODE_PROVIDERS = ["opencode-zen", "opencode-go", "amazon-bedrock", "openai"];
+var OPENCODE_PROVIDERS = ["opencode-zen", "opencode-go", "amazon-bedrock", "openai", "openrouter", "google"];
 var OpenCodeConfigSchema = objectType({
   provider: enumType(OPENCODE_PROVIDERS),
   build_model: stringType().min(1),
@@ -32598,7 +32621,7 @@ var DEFAULT_MODELS = {
   sentinel: "haiku",
   oracle: "opus",
   architect: "opus",
-  keymaker: "haiku"
+  keymaker: "sonnet"
 };
 var CLAUDE_AI_OPT_OUTS = {
   disableClaudeAiConnectors: true,
@@ -32692,6 +32715,7 @@ function mergeHarnessConfig(section, basePath) {
     model: config.default_model || DEFAULT_MODELS.default_model
   };
   applyModeThinking(settings, config);
+  applyAutoCompactWindow(settings, config);
   applyBuiltins(settings, CONFLICTING_BUILTINS);
   for (const name of CLAUDE_OPTIONAL_BUILTINS) {
     if (config.disabled_builtins?.includes(name))
@@ -32757,6 +32781,7 @@ function removeAtelierSettings(claudeDir, config) {
       delete content[key];
   }
   applyModeThinking(content, config, true);
+  applyAutoCompactWindow(content, config, true);
   removeBuiltins(content, CONFLICTING_BUILTINS);
   for (const patch of Object.values(OPTIONAL_BUILTINS))
     removeBuiltins(content, patch);
@@ -32784,6 +32809,27 @@ function applyModeThinking(settings, config, remove = false) {
   else
     delete settings.modelSettings;
 }
+var OPUSPLAN_AUTO_COMPACT_WINDOW = 190000;
+function applyAutoCompactWindow(settings, config, remove = false) {
+  const models = claudeModeModels("opusplan");
+  const enabled = !remove && config.default_model === "opusplan";
+  const modelSettings = { ...settings.modelSettings };
+  for (const model of new Set([models.build, models.plan])) {
+    const entry = { ...modelSettings[model] };
+    if (enabled)
+      entry.autoCompactWindow ??= OPUSPLAN_AUTO_COMPACT_WINDOW;
+    else if (entry.autoCompactWindow === OPUSPLAN_AUTO_COMPACT_WINDOW)
+      delete entry.autoCompactWindow;
+    if (Object.keys(entry).length)
+      modelSettings[model] = entry;
+    else
+      delete modelSettings[model];
+  }
+  if (Object.keys(modelSettings).length)
+    settings.modelSettings = modelSettings;
+  else
+    delete settings.modelSettings;
+}
 function applyBuiltins(settings, patch) {
   Object.assign(settings, patch.settings);
   if (patch.env)
@@ -32891,17 +32937,20 @@ var OPENCODE_PROVIDERS2 = [
   { name: "OpenCode Zen", value: "opencode-zen" },
   { name: "OpenCode Go", value: "opencode-go" },
   { name: "Amazon Bedrock", value: "amazon-bedrock" },
-  { name: "OpenAI", value: "openai" }
+  { name: "OpenAI", value: "openai" },
+  { name: "OpenRouter", value: "openrouter" },
+  { name: "Google Gemini", value: "google" }
 ];
 var PROVIDER_MODELS = {
   "opencode-zen": [
+    "gpt-6.1-sol",
     "gpt-6-astra",
     "gpt-6-sol",
     "gpt-6-luna",
     "claude-opus-5-5",
     "claude-fable-5-1",
     "claude-sonnet-5-5",
-    "claude-haiku-4-5",
+    "claude-haiku-5-5",
     "glm-5.3",
     "glm-5.3-flash",
     "kimi-k3",
@@ -32912,36 +32961,83 @@ var PROVIDER_MODELS = {
     "qwen3.8-max",
     "qwen3.8-flash",
     "mimo-v2.6-flash-free",
+    "gemini-4-argon",
     "gemini-3.8-flash",
-    "gemini-3.1-pro"
+    "grok-4.7",
+    "mistral-large-4",
+    "muse-spark-1.3",
+    "muse-wonder-1.0",
+    "inkling-2-small-free",
+    "nemotron-3.5-super-vl-free",
+    "nemotron-3.5-lightning-free"
   ].map((id) => `opencode/${id}`),
   "opencode-go": [
     "glm-5.3",
     "glm-5.3-flash",
+    "kimi-k3.1-pro",
     "kimi-k3",
     "deepseek-v4.1-flash",
     "deepseek-v4-pro",
+    "minimax-m3.1-flash",
     "minimax-m3",
     "qwen3.8-max",
     "qwen3.8-flash",
     "mimo-v2.6-pro",
     "mimo-v2.6-flash",
-    "gpt-6-luna"
+    "gpt-6-luna",
+    "claude-haiku-5-5",
+    "gemini-4-argon",
+    "grok-4.7",
+    "muse-spark-1.3-contributor"
   ].map((id) => `opencode-go/${id}`),
   "amazon-bedrock": [
     "us.anthropic.claude-opus-5-5",
     "us.anthropic.claude-fable-5-1",
     "us.anthropic.claude-sonnet-5-5",
-    "us.anthropic.claude-haiku-4-5-20251001-v1:0"
+    "us.anthropic.claude-haiku-5-5"
   ].map((id) => `amazon-bedrock/${id}`),
-  openai: ["gpt-6-astra", "gpt-6-sol", "gpt-6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna", "gpt-5.3-codex-spark"].map((id) => `openai/${id}`)
+  openai: ["gpt-6.1-sol", "gpt-6.1-terra", "gpt-6-astra", "gpt-6-sol", "gpt-6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna", "gpt-5.3-codex-spark"].map((id) => `openai/${id}`),
+  openrouter: [
+    "anthropic/claude-opus-5.5",
+    "anthropic/claude-fable-5.1",
+    "anthropic/claude-sonnet-5.5",
+    "anthropic/claude-haiku-5.5",
+    "openai/gpt-6.1-sol",
+    "openai/gpt-6.1-terra",
+    "openai/gpt-6-astra",
+    "openai/gpt-6-sol",
+    "openai/gpt-6-luna",
+    "google/gemini-4-argon",
+    "google/gemini-3.8-flash",
+    "z-ai/glm-5.3",
+    "z-ai/glm-5.3-flash",
+    "moonshotai/kimi-k3.1-pro",
+    "moonshotai/kimi-k3",
+    "deepseek/deepseek-v4.1-flash-1017",
+    "deepseek/deepseek-v4.1-flash",
+    "deepseek/deepseek-v4-pro",
+    "x-ai/grok-4.7",
+    "minimax/minimax-m3.1-flash",
+    "minimax/minimax-m3",
+    "qwen/qwen3.8-flash",
+    "xiaomi/mimo-v2.6-pro",
+    "mistralai/mistral-large-4-0",
+    "thinkingmachines/inkling",
+    "thinkingmachines/inkling-small",
+    "nvidia/nemotron-3.5-super-vl",
+    "nvidia/nemotron-3.5-lightning",
+    "meta/muse-spark-1.3",
+    "meta/muse-spark-open",
+    "meta/muse-wonder"
+  ].map((id) => `openrouter/${id}`),
+  google: ["gemini-4-argon", "gemini-3.8-flash", "gemini-3.7-flash", "gemini-3.5-flash-lite"].map((id) => `google/${id}`)
 };
 var DEFAULT_MODELS2 = {
   "opencode-zen": {
     sentinel: "opencode/glm-5.3-flash",
     oracle: "opencode/kimi-k3",
     architect: "opencode/glm-5.3",
-    keymaker: "opencode/glm-5.3-flash",
+    keymaker: "opencode/deepseek-v4.1-flash",
     build: "opencode/deepseek-v4.1-flash",
     plan: "opencode/glm-5.3"
   },
@@ -32949,15 +33045,15 @@ var DEFAULT_MODELS2 = {
     sentinel: "opencode-go/glm-5.3-flash",
     oracle: "opencode-go/mimo-v2.6-pro",
     architect: "opencode-go/glm-5.3",
-    keymaker: "opencode-go/glm-5.3-flash",
+    keymaker: "opencode-go/deepseek-v4.1-flash",
     build: "opencode-go/deepseek-v4.1-flash",
     plan: "opencode-go/glm-5.3"
   },
   "amazon-bedrock": {
-    sentinel: "amazon-bedrock/us.anthropic.claude-haiku-4-5-20251001-v1:0",
+    sentinel: "amazon-bedrock/us.anthropic.claude-haiku-5-5",
     oracle: "amazon-bedrock/us.anthropic.claude-fable-5-1",
     architect: "amazon-bedrock/us.anthropic.claude-opus-5-5",
-    keymaker: "amazon-bedrock/us.anthropic.claude-haiku-4-5-20251001-v1:0",
+    keymaker: "amazon-bedrock/us.anthropic.claude-sonnet-5-5",
     build: "amazon-bedrock/us.anthropic.claude-sonnet-5-5",
     plan: "amazon-bedrock/us.anthropic.claude-opus-5-5"
   },
@@ -32965,9 +33061,25 @@ var DEFAULT_MODELS2 = {
     sentinel: "openai/gpt-6-luna",
     oracle: "openai/gpt-6-astra",
     architect: "openai/gpt-6-astra",
-    keymaker: "openai/gpt-6-luna",
+    keymaker: "openai/gpt-6-sol",
     build: "openai/gpt-6-sol",
     plan: "openai/gpt-6-astra"
+  },
+  openrouter: {
+    sentinel: "openrouter/z-ai/glm-5.3-flash",
+    oracle: "openrouter/moonshotai/kimi-k3.1-pro",
+    architect: "openrouter/z-ai/glm-5.3",
+    keymaker: "openrouter/deepseek/deepseek-v4.1-flash-1017",
+    build: "openrouter/deepseek/deepseek-v4.1-flash-1017",
+    plan: "openrouter/z-ai/glm-5.3"
+  },
+  google: {
+    sentinel: "google/gemini-3.5-flash-lite",
+    oracle: "google/gemini-4-argon",
+    architect: "google/gemini-4-argon",
+    keymaker: "google/gemini-3.8-flash",
+    build: "google/gemini-3.8-flash",
+    plan: "google/gemini-4-argon"
   }
 };
 var COMMAND_MARKER = "<!-- Generated by Atelier -->";
@@ -32996,7 +33108,7 @@ function defaultSection2(provider) {
       template: name,
       name,
       model: defaults[name],
-      thinking: name === "sentinel" || name === "keymaker" ? selectedProvider === "amazon-bedrock" ? "default" : "low" : name === "oracle" && selectedProvider === "opencode-go" ? "on" : name === "architect" && selectedProvider === "openai" ? "xhigh" : "high"
+      thinking: name === "keymaker" ? selectedProvider === "amazon-bedrock" ? "default" : "high" : name === "sentinel" ? "low" : name === "oracle" && selectedProvider === "opencode-go" ? "on" : name === "architect" && selectedProvider === "openai" ? "xhigh" : "high"
     }))
   };
 }
@@ -34425,7 +34537,7 @@ var DEFAULT_MODELS3 = {
   sentinel: "gpt-6-luna",
   oracle: "gpt-6-astra",
   architect: "gpt-6-astra",
-  keymaker: "gpt-6-luna"
+  keymaker: "gpt-6-sol"
 };
 var WRITABLE_AGENTS = ["keymaker"];
 var MANAGED_CODEX_KEYS = [
@@ -34464,7 +34576,7 @@ function defaultSection3() {
       template: name,
       name,
       model: DEFAULT_MODELS3[name],
-      thinking: { sentinel: "low", oracle: "high", architect: "xhigh", keymaker: "low" }[name]
+      thinking: { sentinel: "low", oracle: "high", architect: "xhigh", keymaker: "medium" }[name]
     }))
   };
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@martinffx/atelier",
-  "version": "3.6.0",
+  "version": "3.7.0",
   "description": "Personal development toolkit - spec-driven development, code quality, deep thinking, and ecosystem patterns",
   "type": "module",
   "bin": {

package/skills/atelier-orchestrator/SKILL.md

@@ -126,8 +126,11 @@ choice to review. If any choice exists, use an Inline Plan.
 Delegate bounded, read-only work to a subagent when it would otherwise flood the main context:
 wide codebase research, finding comparable code, or independent review. Ask for a concise,
 file:line-cited summary, and read the key cited files yourself before deciding. Do not delegate
-interviews or design dialogue with the human, lookups of a few files, or stateful debug and
-implementation loops.
+interviews or design dialogue with the human, lookups of a few files, or stateful debug loops.
+
+Delegate implementation too. Once the design is settled, implementation batches go to
+**Keymaker** subagents through **code-subagents**; the main conversation edits code directly
+only for trivial or mechanical changes and debugging loops.
 
 Every subagent launch must explicitly select an agent. Match the work to the descriptions of
 Atelier's agents (Sentinel, Oracle, Architect, Keymaker) and choose one. Never launch without a
@@ -135,6 +138,21 @@ selection or rely on the harness default. Exception: design-sensitive implementa
 harness's general agent. If Atelier's agents are not installed, explicitly choose the closest
 available agent by description and state the choice in the output.
 
+### Parent context
+
+The main conversation holds the human's decisions and is the scarcest context. Quality degrades
+as it grows, compaction loses detail, and under `opusplan` planning silently drops to Sonnet once
+a turn passes 200k tokens. Keep it lean:
+
+- Send work that produces bulky output to subagents and keep only their summaries. That covers
+  broad search, test and build runs, log reading, and implementation batches through
+  code-subagents.
+- Read only the ranges you need, don't re-read files, and don't paste large outputs back.
+- Persist decisions and progress to the plan or spec artifacts, so compaction or a fresh session
+  loses nothing.
+- Compact deliberately at phase boundaries (plan approved, batch done, before re-planning) with
+  `/compact <focus>`, or start a fresh session. Don't wait for the automatic trigger mid-task.
+
 ## Hard Transitions
 
 | After completing... | Next step |

package/skills/code-subagents/SKILL.md

@@ -70,10 +70,10 @@ For shared-tree parallel work:
 ## Agent Selection
 
 Select the agent explicitly for every dispatch (see Delegation in `atelier-orchestrator`). Give
-implementers the design excerpt so they make no design decisions: mechanical file creation should
-match a scaffolder; design-sensitive implementation may use the harness's general agent, the one
-permitted exception to selecting from Atelier's agents. The batch reviewer is a read-only
-critical review.
+implementers the design excerpt so they make no design decisions. Implementers of settled
+designs, mechanical or not, are Keymaker; design-sensitive implementation may use the harness's
+general agent, the one permitted exception to selecting from Atelier's agents. The batch reviewer
+is a read-only critical review.
 
 ---