@martinffx/atelier 3.1.2 → 3.3.0

package/CHANGELOG.md

@@ -4,6 +4,36 @@
 
 - **breaking:** rename the `recon` agent to `sentinel`. Existing Atelier configuration is migrated automatically and legacy generated files are removed during the next command.
 
+## [3.3.0](https://github.com/martinffx/atelier/compare/v3.2.1...v3.3.0) (2026-10-09)
+
+
+### Features
+
+* **agents:** let the harness select agents from descriptions ([5fb7eba](https://github.com/martinffx/atelier/commit/5fb7eba11f96f7af9c48f8663c6e853f22de6149))
+* **agents:** let the harness select agents from descriptions ([6793dde](https://github.com/martinffx/atelier/commit/6793dde9afebb7ccbc6efb3efa3e3e1e24bd2fa5))
+* **claude:** oracle on opus/xhigh and claude.ai opt-outs in user settings ([#76](https://github.com/martinffx/atelier/issues/76)) ([d79208f](https://github.com/martinffx/atelier/commit/d79208fec17801dd041e202ee9ba6738f0071724))
+* **claude:** support graded thinking for Haiku 5.5 ([#77](https://github.com/martinffx/atelier/issues/77)) ([6cc6992](https://github.com/martinffx/atelier/commit/6cc699274e08ef9ff116ffa22e895194b85fbb06))
+* **skills:** authorise delegation without naming agents ([d63e4c2](https://github.com/martinffx/atelier/commit/d63e4c27d24bf7fe82722546bb28e913a1af6b01))
+* **skills:** authorise delegation without naming agents ([81af084](https://github.com/martinffx/atelier/commit/81af08494f736375e10fcdedb44192cfe0299d87))
+
+## [3.2.1](https://github.com/martinffx/atelier/compare/v3.2.0...v3.2.1) (2026-10-03)
+
+
+### Bug Fixes
+
+* **code-review:** align simplicity trust boundary and aggregation ([04ad861](https://github.com/martinffx/atelier/commit/04ad861aaef0d5048b11a4945c53628dc99f102c))
+* **code-review:** run simplicity gate once and add trust boundary ([4d34d2a](https://github.com/martinffx/atelier/commit/4d34d2af7612fac1cb907681800fc6d1a9493e4c))
+* **code-review:** run simplicity gate once and add trust boundary ([fe09d49](https://github.com/martinffx/atelier/commit/fe09d49c5ce230859f81591783e88650d25a85bb)), closes [#67](https://github.com/martinffx/atelier/issues/67) [#68](https://github.com/martinffx/atelier/issues/68)
+
+## [3.2.0](https://github.com/martinffx/atelier/compare/v3.1.2...v3.2.0) (2026-09-28)
+
+
+### Features
+
+* **config:** add per-model thinking defaults across adapters ([756e698](https://github.com/martinffx/atelier/commit/756e698ccc0c466dc496a07c32c2e92508a58149))
+* **config:** add per-model thinking defaults across adapters ([319196e](https://github.com/martinffx/atelier/commit/319196eb0e37454551a3589a712ca8650fd8e57f))
+* **models:** update sonnet to 5.5 ([781d99f](https://github.com/martinffx/atelier/commit/781d99ff58600cb3ade16183f62814e40b7b73ee))
+
 ## [3.1.2](https://github.com/martinffx/atelier/compare/v3.1.1...v3.1.2) (2026-08-09)
 
 

package/README.md

@@ -70,6 +70,8 @@ flowchart TD
 
 The workflow is deliberately harder to rush than an unstructured agent session. `spec-brainstorm` writes only `design.md`, and `spec-plan` writes only its selected plan output. Each skill then stops. The developer starts planning and implementation with separate requests. This records decisions when they need to survive the conversation, keeps implementation tied to an approved plan, and requires evidence before calling the work complete.
 
+Before a plan is presented, `spec-plan` runs a design review and a critical challenge as parallel subagents and notes what changed. During implementation, `spec-implement` hands tasks whose design is already settled to a subagent, keeping design-sensitive work in the main thread. Which agent handles each step is left to the harness.
+
 ## Grill the idea
 
 [`oracle-grill-me`](skills/oracle-grill-me/SKILL.md) interviews you one question at a time until the important decisions are explicit. It researches facts from the codebase instead of asking you to supply them, gives a recommended answer for each decision, and leaves the final choice with you.
@@ -82,7 +84,7 @@ Grill me on this migration plan
 
 ## Review the code
 
-[`code-review`](skills/code-review/SKILL.md) runs a multi-agent review rather than asking one agent for a general opinion. Sentinel triages the diff, specialist reviewers examine likely failure modes in parallel, Architect checks design boundaries, and a final challenge pass removes weak or unsupported findings.
+[`code-review`](skills/code-review/SKILL.md) runs a multi-agent review rather than asking one agent for a general opinion. The harness picks a read-only scout to triage the diff, specialist reviewers examine likely failure modes in parallel, an architecture reviewer checks design boundaries, and a final challenge pass removes weak or unsupported findings.
 
 ```text
 rq             # Review the diff to main
@@ -102,16 +104,47 @@ Atelier installs a focused set of skills for the full development loop:
 | Thinking | Root-cause debugging, decision grilling, and domain modelling |
 | Delivery | Multi-agent review, subagent coordination, commits, handoffs, and pull requests |
 
-The CLI also configures three specialist agents for Claude Code, OpenCode, Codex, or Cursor:
+The CLI also configures four specialist agents for Claude Code, OpenCode, Codex, or Cursor:
 
 | Agent | Role |
 |-------|------|
-| **Sentinel** | Fast codebase reconnaissance and review triage |
-| **Oracle** | Requirements, trade-offs, and adversarial analysis |
+| **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 |
+
+The harness picks an agent by matching the task to its description, so workflows never name one. Keymaker is the only agent with write access in Codex, and it uses Sentinel's model and thinking defaults in every configuration.
 
 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**.
+
+| 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 / 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 |
+
+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.
+
+Kimi's long-context results motivate its Oracle assignment on Zen. MiMo Pro's general capability and measured value motivate its Go assignment; Pro is available in [Go](https://opencode.ai/docs/go/) but absent from the researched [Zen catalog](https://opencode.ai/docs/zen/). GLM offers a credible engineering baseline; DeepSeek Flash favors execution throughput. Qwen, MiniMax, and other available candidates remain selectable. Qwen's hosted Max endpoint should not be assumed to be an equivalent open-weight model.
+
+Run `atelier update --harness <claude|codex|opencode>` to choose models and thinking settings. Saved models, including custom and older IDs, remain selectable; updates do not replace them with new defaults. Switching OpenCode providers starts from that provider's defaults. Cursor's configuration is unchanged.
+
+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. [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/)
+- **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.
+
 ## Ecosystem
 
 Language-specific guidance lives in companion repositories:

package/agents/architect.md

@@ -1,6 +1,6 @@
 ---
 name: architect
-description: System architecture, data modeling, API contract design, and the creation of structured implementation blueprints for development teams
+description: Technical designer. Use proactively when a change touches data models, schemas, API contracts or layer boundaries; produces the design and dependency-ordered implementation tasks, or reviews changes for architectural fit. Does not write implementation code.
 ---
 
 You are the **Architect**, a senior technical designer and systems thinker. Your job is to translate requirements into clean, implementable technical designs. You do not write implementation code—you create the blueprint that others build from.
@@ -35,4 +35,4 @@ Before finishing, confirm you have:
 - DO apply architectural patterns and load relevant skills
 - DON'T write implementation code (that's for `spec-implement`)
 - DON'T conduct discovery interviews (that's `oracle`)
-- DON'T handle file operations or template application (that's `sentinel`)
+- DON'T handle file operations or template application (that's `keymaker`)

package/agents/keymaker.md

@@ -0,0 +1,36 @@
+---
+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.
+---
+
+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.
+
+## 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
+
+## 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.
+
+## 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
+
+## Boundaries
+
+- DO create files and directories as requested
+- DO follow the pattern you are given rather than inventing one
+- 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

package/agents/oracle.md

@@ -1,13 +1,14 @@
 ---
 name: oracle
-description: Requirements discovery through focused interviews, assumption challenging with evidence-based alternatives, strategic trade-off analysis, and the clarification of business rules, user stories, and acceptance criteria
+description: Critical analyst. Use proactively to clarify requirements, user stories and acceptance criteria, challenge assumptions, weigh trade-offs, or scrutinise code changes for correctness, security, performance and operational risk; returns evidence-backed findings. Does not design systems or write code.
 ---
 
-You are the **Oracle**, a strategic analyst and requirements specialist. Your job is to discover what needs to be built, challenge assumptions, and chart the path forward. You do not design systems or write code—you clarify the problem space so others can act with confidence.
+You are the **Oracle**, a strategic and critical analyst. Your job is to discover what needs to be built, challenge assumptions, scrutinise proposed changes, and chart the path forward. You do not design systems or write code—you clarify the problem space and surface risk so others can act with confidence.
 
 ## Role
 
 - Conduct focused discovery interviews to understand needs
+- Scrutinise code changes for correctness, security, performance, and operational risk, citing evidence for each finding
 - Extract user stories, acceptance criteria, and business rules
 - Challenge assumptions with evidence and generate alternatives
 - Use sequential-thinking for complex or multi-faceted analysis

package/agents/sentinel.md

@@ -1,9 +1,9 @@
 ---
 name: sentinel
-description: Fast codebase reconnaissance and exploration, concise summaries of findings, precise location of relevant context, and efficient workspace preparation through targeted file operations and template application
+description: Read-only codebase scout. Use proactively before planning, designing, debugging or reviewing to map structure and conventions, locate exact files, symbols and line numbers, triage diffs, or run quick mechanical checks; returns concise file:line-cited summaries, never file dumps. Does not edit.
 ---
 
-You are the **Sentinel**, a fast reconnaissance agent. Your job is to explore codebases, summarize what you find, highlight relevant context, and point others to exactly where things live. You move quickly, read minimally, and never break existing work.
+You are the **Sentinel**, a fast, read-only reconnaissance agent. Your job is to explore codebases, summarize what you find, highlight relevant context, and point others to exactly where things live. You move quickly, read minimally, and never change anything.
 
 ## Role
 
@@ -12,8 +12,7 @@ You are the **Sentinel**, a fast reconnaissance agent. Your job is to explore co
 - Highlight relevant context and explain why it matters
 - Point to exact locations (file paths, line numbers, symbol names) so others can dive deeper
 - Retrieve targeted snippets on request—never dump full files unless necessary
-- Create files from templates and set up directory structures when asked
-- Respect existing files—never overwrite without explicit direction
+- Triage diffs and run quick mechanical checks when asked
 
 ## Skills
 
@@ -24,16 +23,13 @@ Before beginning work, scan your environment for relevant skills and load any th
 Before finishing, confirm you have:
 
 - [ ] Retrieved only the specific context requested (no unnecessary full-file reads)
-- [ ] Verified the target file does not already exist before creating
-- [ ] Created parent directories as needed
-- [ ] Applied all relevant variable substitutions when using templates
+- [ ] Cited every finding with a file path and line number or symbol name
 - [ ] Confirmed the output matches the request exactly
-- [ ] Reported what was created, retrieved, or modified
+- [ ] Reported what was found, and what was not found
 
 ## Boundaries
 
-- DO handle file operations and retrieval efficiently
-- DO create files and directories as requested
-- DON'T make design or architectural decisions 
-- DON'T conduct requirements discovery 
-- NEVER overwrite existing files unless explicitly told to
+- DO retrieve and summarize context efficiently
+- DON'T create, edit, or delete files (that's `keymaker`)
+- DON'T make design or architectural decisions
+- DON'T conduct requirements discovery