agestra 4.8.4 → 4.10.0

package/commands/idea.md

@@ -1,61 +1,120 @@
----
-description: "Discover improvements by comparing with similar projects and collecting feedback"
-argument-hint: "[topic or project area]"
----
-
-You are executing the `/agestra idea` command.
-
-**Topic:** $ARGUMENTS
-
-## Step 1: Determine topic
-
-If `$ARGUMENTS` is empty, ask the user what area to explore using AskUserQuestion:
-- "What area would you like to find improvements for? (feature area, project aspect, or general)"
-
-## Step 2: Check environment and select mode
-
-Call `environment_check` to determine which providers and modes are available.
-
-- If **no providers are available**: run the `agestra:agestra-ideator` agent directly (Claude only) with the topic as context. The ideator will research similar projects, collect user complaints, build feature comparisons, and generate prioritized recommendations. Skip to presenting the result.
-- If **1+ providers are available**: proceed to 끝장토론 execution below.
-
-## Step 3: Execute 끝장토론
-
-**팀 구성:** `agestra:agestra-moderator` (리더) + `agestra:agestra-ideator` (Claude) + 사용 가능한 외부 AI (gemini, codex, ollama 등)
-
-1. Independent work + initial aggregation:
-
-   a. In parallel:
-      - Spawn the `agestra:agestra-ideator` agent for Claude's independent improvement research.
-        After the agent completes, save Claude's result as a document via `workspace_create_document`:
-        - **title:** `Idea Exploration — claude/ideator`
-        - **metadata:** `{ "Provider": "claude/ideator", "Task": "{topic}", "Mode": "Independent" }`
-        - **content:** The ideator agent's full output.
-      - For each available provider, call `ai_chat` with `save_as_document` to let each AI produce its own document directly:
-        - **save_as_document.title:** `Idea Exploration — {provider}`
-        - **save_as_document.metadata:** `{ "Task": "{topic}", "Mode": "Independent" }`
-        - **prompt:** Use the appropriate prompt from the `agestra-idea` skill — Mode A (existing project) or Mode B (new project) — including the user interview context collected earlier. The prompt must include structured output requirements (Title, Category, Evidence/Why, Effort, Priority) and the user's constraints.
-
-   b. Collect all document IDs.
-
-   c. Spawn the `agestra:agestra-moderator` agent in **Independent Aggregation** mode:
-      - Pass the **document ID list**.
-      - Moderator reads each document via `workspace_read`.
-      - Moderator classifies: consensus suggestions, unique ideas, disputed priorities.
-      - Moderator creates an **aggregated document** via `workspace_create_document`.
-      - The moderator's integrated document becomes the starting document.
-
-2. Document review rounds (no max — until all agree):
-   a. Moderator sends the current document to each AI for review:
-      - Claude: spawn `agestra:agestra-ideator` → analyze document → write section-by-section feedback
-      - Other providers: `agent_debate_turn` with the document as prompt, requesting agree/disagree per section
-   b. Moderator collects all feedback.
-   c. Classify: agree/disagree per section per provider.
-   d. Revise document incorporating disagreement feedback.
-   e. If all providers agree on all sections → consensus reached.
-   f. If not → next round with revised document.
-   g. **Every 10 rounds:** Ask the user via AskUserQuestion whether to continue or stop with current state.
-
-3. Present the final document:
-   - Consensus sections: marked as agreed
-   - Disputed sections: show split positions with each provider's rationale
+---
+description: "Discover improvements by comparing with similar projects and collecting feedback"
+argument-hint: "[topic or project area]"
+---
+
+You are executing the `/agestra idea` command.
+
+**Topic:** $ARGUMENTS
+
+## Step 0: Setup preflight (MANDATORY)
+
+Before anything else, call `setup_status`. If it reports `Current config: not found` (or the path exists but `setup_status` shows 0 detected providers with a missing-config note), **stop this command and run setup first**:
+
+1. Invoke the `agestra:setup` skill or route the user through `/agestra setup` inline — detect providers, ask for selection + locale, call `setup_apply`.
+2. After setup writes the config, resume this `/agestra idea` command **from Step 1**, preserving the original `$ARGUMENTS` topic. Do not ask the user to retype it.
+
+Rationale: Agestra's path resolver uses a single plugin-scoped `providers.config.json` (`$CLAUDE_PLUGIN_ROOT/providers.config.json` or `~/.agestra/providers.config.json`). Without it, auto-detect silently enables whatever is installed — which has caused disabled providers to participate in past runs. Setup is the only sanctioned way to pick the active set.
+
+## Step 1: Determine topic
+
+If `$ARGUMENTS` is empty, ask the user what area to explore using AskUserQuestion:
+- "What area would you like to find improvements for? (feature area, project aspect, or general)"
+
+## Step 2: Check environment and select mode
+
+Call `environment_check` to determine which providers and modes are available.
+
+- If **no providers are available**: run the `agestra:agestra-ideator` agent directly (Claude only) with the topic as context. Skip to presenting the result.
+- If **1+ providers are available**: proceed to 끝장토론 execution below.
+
+Respect the providers list verbatim. A provider marked `Not found` or missing from `environment_check` MUST NOT be invoked (the `enabled:false` opt-out is honored at registration — do not second-guess it).
+
+## Step 3: Execute 끝장토론
+
+**팀 구성:** `agestra:agestra-moderator` (리더) + `agestra:agestra-ideator` (Claude) + `environment_check`가 Available로 보고한 외부 AI.
+
+### 3.1 Individual exploration (parallel, independent)
+
+Every participant produces an independent analysis **without seeing anyone else's work yet**. Each result is saved as a separate document in `individual/`.
+
+In parallel:
+- Spawn the `agestra:agestra-ideator` agent for Claude's independent improvement research. After it completes, save its output via `workspace_create_document` with:
+  - **kind:** `"individual"`
+  - **title:** `Idea Exploration — claude/ideator`
+  - **metadata:** `{ "Provider": "claude/ideator", "Task": "{topic}", "Mode": "Independent", "Round": "0" }`
+- For each available external provider, call `ai_chat` with `save_as_document` using:
+  - **save_as_document.kind:** `"individual"`
+  - **save_as_document.title:** `Idea Exploration — {provider}`
+  - **save_as_document.metadata:** `{ "Task": "{topic}", "Mode": "Independent", "Round": "0" }`
+  - **prompt:** The Mode A or Mode B prompt from the `agestra-idea` skill, including the user-interview context and the structured output requirements (Title, Category, Evidence, Description, Effort, Priority).
+
+Collect every returned Document ID.
+
+### 3.2 Initial aggregation (Round 0 synthesis)
+
+Spawn the `agestra:agestra-moderator` agent in **Independent Aggregation** mode:
+- Pass the individual document ID list.
+- Moderator reads each via `workspace_read`, classifies consensus / unique / disputed ideas.
+- Moderator writes the aggregated document via `workspace_create_document` with:
+  - **kind:** `"debate"`
+  - **title:** `Aggregated Ideas — {topic} — Round 0`
+  - **metadata:** `{ "Task": "{topic}", "Mode": "Aggregated", "Round": "0" }`
+- This aggregated doc is the **working document** that enters the round loop.
+
+### 3.3 Document review rounds (sequential, read-prior-before-comment)
+
+**Goal:** each round lets every provider review the current working doc AND every earlier-in-round comment, so later providers can agree-and-defer instead of re-stating. This keeps the transcript compact and surfaces real disagreement.
+
+Per round `N` (start at 1):
+
+1. **Freeze inputs**: gather the doc IDs the round will build on — the Round 0 aggregated doc plus every Round `N-1` review doc.
+
+2. **Sequential turns** — call providers one at a time (NOT in parallel). Order: external providers alphabetical first, Claude last. For each provider in turn:
+   a. Read every document produced so far in this round (prior participants' Round `N` reviews), by ID, via `workspace_read`. This is what lets a late-turn provider simply endorse an earlier opinion instead of repeating it.
+   b. Send the review prompt with **all** of: the working doc content, the Round 0 aggregated doc, prior reviews from this round, and the structured agree/disagree contract below.
+   c. Save this provider's round review via `workspace_create_document` with:
+      - **kind:** `"debate"`
+      - **title:** `Round {N} Review — {provider}`
+      - **metadata:** `{ "Task": "{topic}", "Round": "{N}", "Participant": "{provider}" }`
+
+3. **Review-prompt contract** (append to every round prompt):
+
+   > You are reviewing the current working document for Round {N}. Other participants (listed below) have already reviewed it this round — read their reviews first. Where you agree with an existing position, say so and move on; do NOT re-state the same argument. Where you disagree, say so and give a concrete alternative.
+   >
+   > End your review with a machine-parseable block:
+   >
+   > ```
+   > <round-votes>
+   >   <item id="IDEA-01" stance="agree"/>
+   >   <item id="IDEA-02" stance="agree-with-note" note="priority should be MEDIUM not HIGH"/>
+   >   <item id="IDEA-03" stance="disagree" reason="conflicts with stated constraint X"/>
+   >   <item id="IDEA-04" stance="defer-to-{earlier-provider}"/>
+   > </round-votes>
+   > ```
+   >
+   > `stance` is a closed set: `agree | agree-with-note | disagree | defer-to-<providerId>`. One `<item>` per idea in the working doc. Use the idea IDs from the working doc verbatim.
+
+4. **Round wrap-up**: After every provider has taken its turn, the moderator revises the working doc:
+   - Apply changes for `agree-with-note` items the moderator judges well-reasoned.
+   - Mark `disagree` items as Disputed with both positions.
+   - Record every `defer-to` so the final doc can show endorsement chains.
+
+5. **Consensus check**:
+   - If every `<round-votes>` block for this round has no `disagree` on any item AND no open `agree-with-note` that wasn't applied → consensus reached. Go to 3.4.
+   - Else → next round with the revised working doc.
+   - **Every 10 rounds**: call `AskUserQuestion` offering `continue`, `stop with current state`, or `escalate split positions`.
+
+### 3.4 Final consensus document
+
+Write the final document via `workspace_create_document` with:
+- **kind:** `"synthesis"`
+- **title:** `Final — {topic} — Consensus`
+- **metadata:** `{ "Task": "{topic}", "Mode": "Final Consensus", "Rounds": "{N}", "Participants": "{comma-separated}" }`
+- **content:** Use the **Final Consensus Format** defined in the `agestra:agestra-moderator` agent's `<Final_Consensus_Format>` section — explicit per-idea stance table, expanded reasoning in prose (not tables-only), a clear "동의 현황" summary so the reader can see at a glance who agreed and who pushed back.
+
+### 3.5 Present to the user
+
+- Start by naming the final document ID and its path.
+- Show the per-idea stance summary in chat.
+- List each disputed item with the dissenter's rationale verbatim.

package/commands/implement.md

@@ -7,6 +7,15 @@ You are executing the `/agestra implement` command.
 
 **Task:** $ARGUMENTS
 
+## Step 0: Setup preflight (MANDATORY)
+
+Before anything else, call `setup_status`. If it reports `Current config: not found`, **stop this command and run setup first**:
+
+1. Invoke the `agestra:setup` skill (or `/agestra setup` inline) — provider detection, selection, locale, `setup_apply`.
+2. After the config is written, resume this `/agestra implement` command **from Step 1**, preserving `$ARGUMENTS`. Do not ask the user to retype.
+
+Agestra uses a single plugin-scoped `providers.config.json`. No config → no sanctioned provider set → setup is the only correct starting point.
+
 ## Step 1: Determine implementation target
 
 If `$ARGUMENTS` is empty, ask the user:

package/commands/review.md

@@ -1,111 +1,120 @@
----
-description: "Review code quality, security, and integration completeness"
-argument-hint: "[target file, directory, or description]"
----
-
-You are executing the `/agestra review` command.
-
-**Target:** $ARGUMENTS
-
-## Step 1: Determine review scope
-
-If `$ARGUMENTS` is provided, use it as the review target and skip the scope question.
-
-If `$ARGUMENTS` is empty, ask the user using AskUserQuestion (in the user's language):
-
-| Option | Description |
-|--------|-------------|
-| **전체 프로젝트** | 프로젝트 전체를 리뷰 |
-| **일부 지정** | 특정 파일, 디렉토리, 또는 영역을 지정하여 리뷰 |
-
-- If **"전체 프로젝트"**: set target to the project root.
-- If **"일부 지정"**: ask a follow-up "What would you like to review? (file path, directory, or description)" and use the answer as the target.
-
-## Step 2: Choose review focus areas
-
-Ask the user what to focus on using AskUserQuestion with **multiSelect: true** (in the user's language):
-
-| Option | Description |
-|--------|-------------|
-| **전체 점검** | 아래 모든 항목을 점검 |
-| **보안 취약점** | OWASP top 10 (인젝션, 인증, XSS 등) |
-| **하드코딩** | 매직넘버, 하드코딩된 URL, 임베디드 자격증명 |
-| **중복 코드** | 반복되는 로직, 복사-붙여넣기 코드 |
-| **레거시 코드** | 더 이상 사용되지 않는 코드, 고아 시스템, 데드 코드 |
-| **스파게티 코드** | 과도한 복잡성, 긴 함수, 깊은 중첩, 얽힌 의존성 |
-| **테스트 커버리지** | 테스트 없는 공개 함수, 엣지케이스 미커버 |
-| **i18n** | 번역 함수 없이 하드코딩된 UI 문자열 (i18n 시스템이 있는 경우만) |
-| **스펙 불일치** | 설계 문서와 실제 구현의 차이 |
-
-- If **"전체 점검"** is selected: use all focus areas.
-- If specific items are selected: use only those areas.
-- If **"Other"**: use the user's custom focus description.
-
-Pass the selected focus areas to all reviewers (agent and external AIs) as part of the review instructions.
-
-## Step 3: Check environment and select mode
-
-Call `environment_check` to determine which providers and modes are available.
-
-- If **no providers are available**: run the `agestra:agestra-reviewer` agent directly (Claude only) with the target and selected focus areas as context. Skip to presenting the result.
-- If **1+ providers are available**: proceed to 끝장토론 execution below.
-
-## Step 4: Execute 끝장토론
-
-### Execution discipline
-
-- Prefer Agestra MCP tools for external providers (`ai_chat`, `ai_analyze_files`, `agent_debate_*`, `workspace_*`, `cli_job_submit` when a direct provider retry is needed).
-- Do NOT launch raw Bash retries such as `codex exec ...` or `gemini ...` yourself, and do NOT invent provider-specific flags. If a retry is required, use `cli_job_submit` with the provider's default mapping unless you have just verified the current CLI help output in this environment.
-- Exclude `ollama` from review workflows even if it is available. For repository/code reviews, use only review-capable external providers such as `gemini`, `codex`, and registered Claude-backed reviewers.
-- Treat the Claude reviewer agent as asynchronous work that may legitimately take several minutes. Poll about once per minute; an empty or slowly growing output file is not a failure by itself.
-- Do NOT stop or replace the Claude reviewer with a duplicate main-thread review unless there is an explicit error, the user asks to cancel, or there has been no visible progress for at least 8 minutes. For large review scopes, allow up to 15 minutes before declaring the reviewer stalled.
-- If a background reviewer is still running, tell the user you are waiting and continue the orchestration. Do not short-circuit the workflow just because another provider finished earlier.
-- Use the turn-based flow (`agent_debate_create` + iterative `agent_debate_review` / `agent_debate_turn` + `agent_debate_conclude`) or the approval-gated flow (`agent_debate_structured` + `agent_debate_approve`/`_continue`/`_reject`) so long-running review rounds do not get cut off by host tool-call time limits.
-
-**팀 구성:** `agestra:agestra-moderator` (리더) + `agestra:agestra-reviewer` (Claude) + 리뷰용 외부 AI (`gemini`, `codex`, 등록된 Claude-backed reviewer 등; `ollama` 제외)
-
-1. Independent work + initial aggregation:
-
-   a. In parallel:
-      - Spawn the `agestra:agestra-reviewer` agent for Claude's independent analysis.
-        After the agent completes, save Claude's result as a document via `workspace_create_document`:
-        - **title:** `Code Review — claude/reviewer`
-        - **metadata:** `{ "Provider": "claude/reviewer", "Task": "{review target}", "Focus": "{selected focus areas}", "Mode": "Independent" }`
-        - **content:** The reviewer agent's full output.
-      - For each available provider, call `ai_chat` with `save_as_document` to let each AI produce its own document directly:
-        - **save_as_document.title:** `Code Review — {provider}`
-        - **save_as_document.metadata:** `{ "Task": "{review target}", "Focus": "{selected focus areas}", "Mode": "Independent" }`
-        - **prompt:**
-
-          > Review the following code. Focus on: [selected focus areas].
-          > For each finding, provide severity (CRITICAL/HIGH/MEDIUM/LOW), file:line location, and evidence.
-          >
-          > Target: [the review target]
-
-   b. Collect all document IDs.
-
-   c. Spawn the `agestra:agestra-moderator` agent in **Independent Aggregation** mode:
-      - Pass the **document ID list**.
-      - Moderator reads each document via `workspace_read`.
-      - Moderator classifies: consensus findings, unique findings, disputed points.
-      - Moderator creates an **aggregated document** via `workspace_create_document`.
-      - The moderator's integrated document becomes the starting document.
-
-2. Start a turn-based debate session with `agent_debate_create`.
-   - Use the integrated document's title/topic as the debate topic.
-   - Reuse the same reviewer set from step 1, excluding `ollama`.
-
-3. Document review rounds (no max — until all agree):
-   a. Moderator sends the current document to Claude reviewer asynchronously.
-   b. Moderator sends the same current document to non-Claude providers with `agent_debate_review`.
-   c. Moderator collects all feedback and updates the working document.
-   d. If a provider needs a follow-up reply to another provider's point, add a targeted `agent_debate_turn` in the same debate session rather than restarting the workflow.
-   e. If all providers agree on all sections → consensus reached.
-   f. If not → next round with the revised document.
-   g. **Every 10 rounds:** Ask the user via AskUserQuestion whether to continue or stop with current state.
-
-4. When consensus is reached, record the moderator's final synthesis with `agent_debate_conclude`.
-
-5. Present the final document:
-   - Consensus sections: marked as agreed
-   - Disputed sections: show split positions with each provider's rationale
+---
+description: "Review code quality, security, and integration completeness"
+argument-hint: "[target file, directory, or description]"
+---
+
+You are executing the `/agestra review` command.
+
+**Target:** $ARGUMENTS
+
+## Step 0: Setup preflight (MANDATORY)
+
+Before anything else, call `setup_status`. If it reports `Current config: not found`, **stop this command and run setup first**:
+
+1. Invoke the `agestra:setup` skill (or `/agestra setup` inline) — provider detection, selection, locale, `setup_apply`.
+2. After the config is written, resume this `/agestra review` command **from Step 1**, preserving `$ARGUMENTS`. Do not ask the user to retype.
+
+Agestra uses a single plugin-scoped `providers.config.json`. No config → no sanctioned provider set → setup is the only correct starting point.
+
+## Step 1: Determine review scope
+
+If `$ARGUMENTS` is provided, use it as the review target and skip the scope question.
+
+If `$ARGUMENTS` is empty, ask the user using AskUserQuestion (in the user's language):
+
+| Option | Description |
+|--------|-------------|
+| **전체 프로젝트** | 프로젝트 전체를 리뷰 |
+| **일부 지정** | 특정 파일, 디렉토리, 또는 영역을 지정하여 리뷰 |
+
+- If **"전체 프로젝트"**: set target to the project root.
+- If **"일부 지정"**: ask a follow-up "What would you like to review? (file path, directory, or description)" and use the answer as the target.
+
+## Step 2: Choose review focus areas
+
+Ask the user what to focus on using AskUserQuestion with **multiSelect: true** (in the user's language):
+
+| Option | Description |
+|--------|-------------|
+| **전체 점검** | 아래 모든 항목을 점검 |
+| **보안 취약점** | OWASP top 10 (인젝션, 인증, XSS 등) |
+| **하드코딩** | 매직넘버, 하드코딩된 URL, 임베디드 자격증명 |
+| **중복 코드** | 반복되는 로직, 복사-붙여넣기 코드 |
+| **레거시 코드** | 더 이상 사용되지 않는 코드, 고아 시스템, 데드 코드 |
+| **스파게티 코드** | 과도한 복잡성, 긴 함수, 깊은 중첩, 얽힌 의존성 |
+| **테스트 커버리지** | 테스트 없는 공개 함수, 엣지케이스 미커버 |
+| **i18n** | 번역 함수 없이 하드코딩된 UI 문자열 (i18n 시스템이 있는 경우만) |
+| **스펙 불일치** | 설계 문서와 실제 구현의 차이 |
+
+- If **"전체 점검"** is selected: use all focus areas.
+- If specific items are selected: use only those areas.
+- If **"Other"**: use the user's custom focus description.
+
+Pass the selected focus areas to all reviewers (agent and external AIs) as part of the review instructions.
+
+## Step 3: Check environment and select mode
+
+Call `environment_check` to determine which providers and modes are available.
+
+- If **no providers are available**: run the `agestra:agestra-reviewer` agent directly (Claude only) with the target and selected focus areas as context. Skip to presenting the result.
+- If **1+ providers are available**: proceed to 끝장토론 execution below.
+
+## Step 4: Execute 끝장토론
+
+### Execution discipline
+
+- Prefer Agestra MCP tools for external providers (`ai_chat`, `ai_analyze_files`, `agent_debate_*`, `workspace_*`, `cli_job_submit` when a direct provider retry is needed).
+- Do NOT launch raw Bash retries such as `codex exec ...` or `gemini ...` yourself, and do NOT invent provider-specific flags. If a retry is required, use `cli_job_submit` with the provider's default mapping unless you have just verified the current CLI help output in this environment.
+- Exclude `ollama` from review workflows even if it is available. For repository/code reviews, use only review-capable external providers such as `gemini`, `codex`, and registered Claude-backed reviewers.
+- Treat the Claude reviewer agent as asynchronous work that may legitimately take several minutes. Poll about once per minute; an empty or slowly growing output file is not a failure by itself.
+- Do NOT stop or replace the Claude reviewer with a duplicate main-thread review unless there is an explicit error, the user asks to cancel, or there has been no visible progress for at least 8 minutes. For large review scopes, allow up to 15 minutes before declaring the reviewer stalled.
+- If a background reviewer is still running, tell the user you are waiting and continue the orchestration. Do not short-circuit the workflow just because another provider finished earlier.
+- Use the turn-based flow (`agent_debate_create` + iterative `agent_debate_review` / `agent_debate_turn` + `agent_debate_conclude`) or the approval-gated flow (`agent_debate_structured` + `agent_debate_approve`/`_continue`/`_reject`) so long-running review rounds do not get cut off by host tool-call time limits.
+
+**팀 구성:** `agestra:agestra-moderator` (리더) + `agestra:agestra-reviewer` (Claude) + 리뷰용 외부 AI (`gemini`, `codex`, 등록된 Claude-backed reviewer 등; `ollama` 제외)
+
+1. Independent work + initial aggregation:
+
+   a. In parallel:
+      - Spawn the `agestra:agestra-reviewer` agent for Claude's independent analysis.
+        After the agent completes, save Claude's result as a document via `workspace_create_document`:
+        - **title:** `Code Review — claude/reviewer`
+        - **metadata:** `{ "Provider": "claude/reviewer", "Task": "{review target}", "Focus": "{selected focus areas}", "Mode": "Independent" }`
+        - **content:** The reviewer agent's full output.
+      - For each available provider, call `ai_chat` with `save_as_document` to let each AI produce its own document directly:
+        - **save_as_document.title:** `Code Review — {provider}`
+        - **save_as_document.metadata:** `{ "Task": "{review target}", "Focus": "{selected focus areas}", "Mode": "Independent" }`
+        - **prompt:**
+
+          > Review the following code. Focus on: [selected focus areas].
+          > For each finding, provide severity (CRITICAL/HIGH/MEDIUM/LOW), file:line location, and evidence.
+          >
+          > Target: [the review target]
+
+   b. Collect all document IDs.
+
+   c. Spawn the `agestra:agestra-moderator` agent in **Independent Aggregation** mode:
+      - Pass the **document ID list**.
+      - Moderator reads each document via `workspace_read`.
+      - Moderator classifies: consensus findings, unique findings, disputed points.
+      - Moderator creates an **aggregated document** via `workspace_create_document`.
+      - The moderator's integrated document becomes the starting document.
+
+2. Start a turn-based debate session with `agent_debate_create`.
+   - Use the integrated document's title/topic as the debate topic.
+   - Reuse the same reviewer set from step 1, excluding `ollama`.
+
+3. Document review rounds (no max — until all agree):
+   a. Moderator sends the current document to Claude reviewer asynchronously.
+   b. Moderator sends the same current document to non-Claude providers with `agent_debate_review`.
+   c. Moderator collects all feedback and updates the working document.
+   d. If a provider needs a follow-up reply to another provider's point, add a targeted `agent_debate_turn` in the same debate session rather than restarting the workflow.
+   e. If all providers agree on all sections → consensus reached.
+   f. If not → next round with the revised document.
+   g. **Every 10 rounds:** Ask the user via AskUserQuestion whether to continue or stop with current state.
+
+4. When consensus is reached, record the moderator's final synthesis with `agent_debate_conclude`.
+
+5. Present the final document:
+   - Consensus sections: marked as agreed
+   - Disputed sections: show split positions with each provider's rationale

package/commands/setup.md

@@ -0,0 +1,71 @@
+---
+description: "Select AI providers and UI language for Agestra workflows"
+argument-hint: ""
+---
+
+You are executing the `/agestra setup` command.
+
+## Step 1: Inspect current state
+
+Call these tools first:
+- `environment_check`
+- `provider_list`
+- `setup_status`
+
+Use the results to identify:
+- which providers are currently available
+- which providers are already enabled in `providers.config.json`
+- which locale is currently configured
+
+## Step 2: Handle the no-provider case
+
+If `setup_status` shows that no providers are detected:
+- Explain this in the user's language.
+- Briefly tell the user they need to install at least one supported provider (Ollama, Gemini CLI, Codex CLI, or Claude CLI).
+- Stop without calling `setup_apply`.
+
+## Step 3: Ask for provider selection
+
+Use AskUserQuestion with **multiSelect: true** in the user's language.
+
+Present one option per currently available provider from `setup_status`.
+- Label: provider ID (`ollama`, `gemini`, `codex`, `claude-cli`, etc.)
+- Description: short capability hint from `provider_list`
+
+Guidance:
+- If one or more providers are already enabled, pre-select them conceptually in your reasoning.
+- Recommend enabling at least one external provider.
+- If the user wants Claude-only operation, allow an empty selection only if they explicitly ask for it. Otherwise prefer at least one enabled provider.
+
+## Step 4: Ask for language
+
+Use AskUserQuestion in the user's language with these choices:
+
+| Option | Description |
+|--------|-------------|
+| **한국어** | Korean UI / moderator text |
+| **English** | English UI / moderator text |
+| **日本語** | Japanese UI / moderator text |
+| **中文** | Chinese UI / moderator text |
+
+Map these to locale codes:
+- `한국어` → `ko`
+- `English` → `en`
+- `日本語` → `ja`
+- `中文` → `zh`
+
+## Step 5: Apply setup
+
+Call `setup_apply` with:
+- `enabled_providers`: the selected provider IDs
+- `locale`: the selected locale code
+- `selection_policy`: `default-only`
+
+## Step 6: Report result
+
+Respond in the user's language with:
+- enabled providers
+- selected language
+- the path to `providers.config.json`
+
+If the user changed providers, mention that future Agestra workflows will use the new setup on the next run/read of config.