maxsimcli 5.0.6 → 5.1.0

package/dist/assets/templates/commands/maxsim/debug.md

@@ -1,167 +1,33 @@
 ---
 name: maxsim:debug
-description: Systematic debugging with persistent state across context resets
-argument-hint: [issue description]
-allowed-tools:
-  - Read
-  - Bash
-  - Task
-  - AskUserQuestion
+description: Systematic debugging with reproduce-hypothesize-isolate-verify-fix cycle
+argument-hint: "[issue description]"
+allowed-tools: [Read, Write, Edit, Bash, Grep, Glob, Agent, AskUserQuestion]
 ---
 
 <objective>
-Debug issues using scientific method with subagent isolation.
-
-**Orchestrator role:** Gather symptoms, spawn verifier agent (debug mode), handle checkpoints, spawn continuations.
-
-**Why subagent:** Investigation burns context fast (reading files, forming hypotheses, testing). Fresh 200k context per investigation. Main context stays lean for user interaction.
+Debug issues using a scientific method cycle: Reproduce → Hypothesize → Isolate → Verify → Fix. Active debug sessions are tracked as GitHub Issues (label: `debug`).
 </objective>
 
 <context>
-User's issue: $ARGUMENTS
+Arguments: $ARGUMENTS
+
+If $ARGUMENTS is provided, treat it as the issue description and skip straight to symptom confirmation.
 
-Check for active sessions:
-```bash
-ls .planning/debug/*.md 2>/dev/null | grep -v resolved | head -5
-```
+GitHub is the sole source of truth. Check for open Issues labeled `debug` to detect active sessions — no .planning/debug/ files.
 </context>
 
 <process>
-
-## 0. Initialize Context
-
-```bash
-INIT=$(node ~/.claude/maxsim/bin/maxsim-tools.cjs state load)
-```
-
-Extract `commit_docs` from init JSON. Resolve debugger model:
-```bash
-DEBUGGER_MODEL=$(node ~/.claude/maxsim/bin/maxsim-tools.cjs resolve-model verifier --raw)
-```
-
-## 1. Check Active Sessions
-
-If active sessions exist AND no $ARGUMENTS:
-- List sessions with status, hypothesis, next action
-- User picks number to resume OR describes new issue
-
-If $ARGUMENTS provided OR user describes new issue:
-- Continue to symptom gathering
-
-## 2. Gather Symptoms (if new issue)
-
-Use AskUserQuestion for each:
-
-1. **Expected behavior** - What should happen?
-2. **Actual behavior** - What happens instead?
-3. **Error messages** - Any errors? (paste or describe)
-4. **Timeline** - When did this start? Ever worked?
-5. **Reproduction** - How do you trigger it?
-
-After all gathered, confirm ready to investigate.
-
-## 3. Spawn Verifier Agent (Debug Mode)
-
-Fill prompt and spawn:
-
-```markdown
-<objective>
-Investigate issue: {slug}
-
-**Summary:** {trigger}
-</objective>
-
-<symptoms>
-expected: {expected}
-actual: {actual}
-errors: {errors}
-reproduction: {reproduction}
-timeline: {timeline}
-</symptoms>
-
-<mode>
-symptoms_prefilled: true
-goal: find_and_fix
-</mode>
-
-<debug_file>
-Create: .planning/debug/{slug}.md
-</debug_file>
-```
-
-```
-Task(
-  prompt=filled_prompt,
-  subagent_type="verifier",
-  model="{debugger_model}",
-  description="Debug {slug}"
-)
-```
-
-## 4. Handle Agent Return
-
-**If `## ROOT CAUSE FOUND`:**
-- Display root cause and evidence summary
-- Offer options:
-  - "Fix now" - spawn fix subagent
-  - "Plan fix" - suggest /maxsim:plan --gaps
-  - "Manual fix" - done
-
-**If `## CHECKPOINT REACHED`:**
-- Present checkpoint details to user
-- Get user response
-- If checkpoint type is `human-verify`:
-  - If user confirms fixed: continue so agent can finalize/resolve/archive
-  - If user reports issues: continue so agent returns to investigation/fixing
-- Spawn continuation agent (see step 5)
-
-**If `## INVESTIGATION INCONCLUSIVE`:**
-- Show what was checked and eliminated
-- Offer options:
-  - "Continue investigating" - spawn new agent with additional context
-  - "Manual investigation" - done
-  - "Add more context" - gather more symptoms, spawn again
-
-## 5. Spawn Continuation Agent (After Checkpoint)
-
-When user responds to checkpoint, spawn fresh agent:
-
-```markdown
-<objective>
-Continue debugging {slug}. Evidence is in the debug file.
-</objective>
-
-<prior_state>
-<files_to_read>
-- .planning/debug/{slug}.md (Debug session state)
-</files_to_read>
-</prior_state>
-
-<checkpoint_response>
-**Type:** {checkpoint_type}
-**Response:** {user_response}
-</checkpoint_response>
-
-<mode>
-goal: find_and_fix
-</mode>
-```
-
-```
-Task(
-  prompt=continuation_prompt,
-  subagent_type="verifier",
-  model="{debugger_model}",
-  description="Continue debug {slug}"
-)
-```
-
+Follow @~/.claude/maxsim/workflows/debug.md end-to-end.
+
+1. Check GitHub for open Issues labeled `debug` (active sessions)
+   - If active sessions exist and no $ARGUMENTS: list them, let user pick to resume or start new
+2. Gather symptoms via AskUserQuestion: expected behavior, actual behavior, errors, timeline, reproduction steps
+3. Create a GitHub Issue labeled `debug` to track the session
+4. Spawn a verifier Agent with symptoms and the GitHub Issue URL as context
+5. Handle Agent return:
+   - Root cause found → display findings, offer Fix / Plan Fix / Manual
+   - Checkpoint reached → surface to user, get response, spawn continuation Agent
+   - Inconclusive → show what was eliminated, offer Continue / Manual / Add Context
+6. On fix: close the GitHub debug Issue with resolution summary
 </process>
-
-<success_criteria>
-- [ ] Active sessions checked
-- [ ] Symptoms gathered (if new)
-- [ ] Verifier agent spawned with debug context
-- [ ] Checkpoints handled correctly
-- [ ] Root cause confirmed before fixing
-</success_criteria>

package/dist/assets/templates/commands/maxsim/execute.md

@@ -1,45 +1,31 @@
 ---
 name: maxsim:execute
-description: Execute all plans in a phase with auto-verification and retry
-argument-hint: "<phase-number> [--worktrees|--no-worktrees]"
-allowed-tools:
-  - Read
-  - Write
-  - Edit
-  - Glob
-  - Grep
-  - Bash
-  - Task
-  - TodoWrite
+description: Execute all plans in a phase with parallel agents and auto-verification
+argument-hint: "<phase-number>"
+allowed-tools: [Read, Write, Edit, Bash, Grep, Glob, Agent]
 ---
-<objective>
-Execute the phase state machine: Execute all plans in wave order, auto-verify, retry on failure (max 2 retries).
-
-**How it works:**
-1. Detect phase state (already done, partially executed, ready to execute)
-2. Execute all plans grouped by wave — parallel within waves, sequential across waves
-3. Auto-verify after all plans complete — spawn verifier agent
-4. If verification fails, auto-retry with gap closure (max 2 retries, 3 total attempts)
-5. On final failure, report what failed and let user decide
-6. Supports worktree-based parallel execution: --worktrees forces worktree isolation, --no-worktrees forces standard mode
 
-**Re-entry:** If phase is already executed and verified, show status and offer options (view results, re-execute, view verification).
-
-**Phase-level only:** Operates on the entire phase — no plan-level granularity.
+<objective>
+Execute all task Issues in a phase using parallel Agents, auto-verify results, and retry on failure (max 2 retries). GitHub Issues track task state throughout.
 </objective>
 
-<execution_context>
-@~/.claude/maxsim/workflows/execute.md
-@~/.claude/maxsim/references/ui-brand.md
-</execution_context>
-
 <context>
-Phase number: $ARGUMENTS (required — e.g., `/maxsim:execute 3`)
+Arguments: $ARGUMENTS
+
+Phase number is required (e.g., `/maxsim:execute 3`).
 
-Context files are resolved inside the workflow via `maxsim-tools init execute-phase` and per-subagent context assembly.
+GitHub is the sole source of truth. Task state (todo/in-progress/done) is read from GitHub Issue labels and Project Board columns — no .planning/ files.
+
+Re-entry: If phase is already executed and verified, show status and offer options (view results, re-execute).
 </context>
 
 <process>
-Execute the execute workflow from @~/.claude/maxsim/workflows/execute.md end-to-end.
-Preserve all workflow gates (state detection, wave execution, verification, retry loop, re-entry flow).
+Follow @~/.claude/maxsim/workflows/execute.md end-to-end.
+
+1. Detect phase state from GitHub Project Board (already done / partially executed / ready)
+2. Group task Issues by wave — execute parallel Agents within each wave, sequential across waves
+3. Each Agent updates its GitHub Issue label on start and completion
+4. After all tasks complete, spawn a verifier Agent to check acceptance criteria
+5. On verification failure, auto-retry with gap-closure context (max 2 retries)
+6. On final failure, report what failed and surface options to user
 </process>

package/dist/assets/templates/commands/maxsim/go.md

@@ -1,29 +1,30 @@
 ---
 name: maxsim:go
-description: Auto-detect project state and dispatch to the right command
-allowed-tools:
-  - Read
-  - Bash
-  - Glob
-  - Grep
-  - SlashCommand
+description: Auto-detect project state and execute the right action
+allowed-tools: [Read, Bash, Grep, Glob, Agent, AskUserQuestion]
 ---
-<objective>
-Auto-detect project state through deep context gathering, surface any problems, and dispatch to the appropriate MAXSIM command.
-
-**How it works:**
-1. Gather deep context (project state, git status, recent commits, blockers)
-2. Surface any problems and block until resolved
-3. Show detection reasoning (what was found)
-4. Act immediately by dispatching to the right command
 
-Show + Act pattern: display detection reasoning, then act. No arguments -- pure auto-detection. User can Ctrl+C if the detection is wrong.
+<objective>
+Auto-detect project state from GitHub and dispatch to the right action. No arguments — pure auto-detection. Show detection reasoning, enter Plan Mode, propose the action, then execute after user approval.
 </objective>
 
-<execution_context>
-@~/.claude/maxsim/workflows/go.md
-</execution_context>
+<context>
+GitHub is the sole source of truth. Read the GitHub Project Board, open Issues, and Milestones to determine current state. Do NOT read .planning/ files.
+
+Detection priority:
+1. No CLAUDE.md or GitHub Milestone → init needed
+2. Active Milestone with no planned phase → plan next phase
+3. Active Milestone with planned phase not executed → execute that phase
+4. Open bug/issue labeled `bug` → debug workflow
+5. All phases done → offer milestone completion
+</context>
 
 <process>
-Execute the go workflow from @~/.claude/maxsim/workflows/go.md end-to-end.
+Follow @~/.claude/maxsim/workflows/go.md end-to-end.
+
+1. Read GitHub Project Board state via `gh` CLI
+2. Detect what's next using the priority list above
+3. Enter Plan Mode — show detection reasoning and proposed action
+4. Wait for user approval (Ctrl+C cancels)
+5. Execute approved action by spawning the appropriate Agent
 </process>

package/dist/assets/templates/commands/maxsim/help.md

@@ -1,22 +1,13 @@
 ---
 name: maxsim:help
-description: Show MAXSIM command reference and usage guide
+description: Show available MaxsimCLI commands and usage
+allowed-tools: [Read]
 ---
-<objective>
-Display the complete MAXSIM command reference.
 
-Output ONLY the reference content below. Do NOT add:
-- Project-specific analysis
-- Git status or file context
-- Next-step suggestions
-- Any commentary beyond the reference
+<objective>
+Display the complete MaxsimCLI command reference. Output ONLY the reference content — no project analysis, no git status, no commentary.
 </objective>
 
-<execution_context>
-@~/.claude/maxsim/workflows/help.md
-</execution_context>
-
 <process>
-Output the complete MAXSIM command reference from @~/.claude/maxsim/workflows/help.md.
-Display directly -- no additions or modifications.
+Output the command reference from @~/.claude/maxsim/workflows/help.md directly. No additions or modifications.
 </process>

package/dist/assets/templates/commands/maxsim/init.md

@@ -1,52 +1,30 @@
 ---
 name: maxsim:init
-description: Initialize a new project or manage milestone lifecycle
+description: Initialize MaxsimCLI in current project with GitHub integration
 argument-hint: "[--auto]"
-allowed-tools:
-  - Read
-  - Bash
-  - Write
-  - Task
-  - AskUserQuestion
-  - Glob
-  - Grep
-  - WebFetch
-  - mcp__context7__resolve-library-id
-  - mcp__context7__get-library-docs
+allowed-tools: [Read, Write, Edit, Bash, Grep, Glob, Agent, AskUserQuestion, WebFetch]
 ---
-<context>
-**Flags:**
-- `--auto` -- Automatic mode. After config questions, runs the appropriate flow without further interaction. For new projects, expects idea document via @ reference.
-</context>
 
 <objective>
-Unified project initialization. Detects whether this is a new project, existing project, or milestone lifecycle and routes to the appropriate workflow.
+Initialize MaxsimCLI for the current project. Scans the repo, interviews the user, sets up GitHub Milestone and Project Board, writes CLAUDE.md, and optionally generates a Roadmap.
+</objective>
 
-**Creates (depending on scenario):**
-- `.planning/PROJECT.md` -- project context
-- `.planning/config.json` -- workflow preferences
-- `.planning/REQUIREMENTS.md` -- scoped requirements
-- `.planning/ROADMAP.md` -- phase structure
-- `.planning/STATE.md` -- project memory
-- `.planning/DECISIONS.md` -- key decisions with rationale
-- `.planning/ACCEPTANCE-CRITERIA.md` -- measurable success criteria
-- `.planning/NO-GOS.md` -- explicit exclusions and anti-patterns
-- `.planning/codebase/` -- codebase analysis (existing projects only)
-- `.planning/research/` -- domain research (optional)
+<context>
+Arguments: $ARGUMENTS
 
-**After this command:** Run `/maxsim:plan 1` to start phase planning.
-</objective>
+Flags:
+- `--auto` — Skip confirmations after config questions. For new projects, expects an idea document via @ reference.
 
-<execution_context>
-@~/.claude/maxsim/workflows/init.md
-@~/.claude/maxsim/references/questioning.md
-@~/.claude/maxsim/references/thinking-partner.md
-@~/.claude/maxsim/references/ui-brand.md
-@~/.claude/maxsim/templates/project.md
-@~/.claude/maxsim/templates/requirements.md
-</execution_context>
+GitHub is the sole source of truth. Init creates GitHub Milestones and Issues — no .planning/ files.
+</context>
 
 <process>
-Execute the init workflow from @~/.claude/maxsim/workflows/init.md end-to-end.
-Pass $ARGUMENTS through to the workflow for flag handling (--auto).
+Follow @~/.claude/maxsim/workflows/init.md end-to-end.
+
+1. Scan repo structure (language, framework, existing CI)
+2. Interview user: project goal, scope, constraints, success criteria
+3. Create GitHub Milestone for this project/version
+4. Write CLAUDE.md with project context and MaxsimCLI config
+5. Optionally generate Roadmap as GitHub Issues (phases)
+6. Confirm setup and suggest `/maxsim:plan 1` as next step
 </process>

package/dist/assets/templates/commands/maxsim/plan.md

@@ -1,50 +1,35 @@
 ---
 name: maxsim:plan
-description: State-machine plan command — Discussion, Research, Planning stages with gate confirmations
-argument-hint: "[phase-number] [--force-research] [--skip-verify]"
-allowed-tools:
-  - Read
-  - Write
-  - Bash
-  - Glob
-  - Grep
-  - Task
-  - WebFetch
-  - mcp__context7__*
+description: Plan a specific phase with discussion, research, and task breakdown
+argument-hint: "<phase-number> [--force-research] [--skip-verify]"
+allowed-tools: [Read, Write, Bash, Grep, Glob, Agent, AskUserQuestion, WebFetch]
 ---
-<objective>
-Execute the plan state machine: Discussion -> Research -> Planning. Each stage produces artifacts, shows a gate summary, and waits for user confirmation before advancing.
-
-**How it works:**
-1. Detect current stage from phase artifacts (CONTEXT.md, RESEARCH.md, PLAN.md)
-2. Start at the earliest incomplete stage
-3. Run each stage via dedicated sub-workflow
-4. Show rich gate summary after each stage — wait for user confirmation
-5. Advance to next stage on confirmation
 
-**Re-entry:** If phase is already planned, show status and offer options (view, re-plan, execute).
-
-**Flags:**
-- `--force-research` — Re-run research even if RESEARCH.md exists
-- `--skip-verify` — Skip plan verification loop after planning stage
+<objective>
+Run the plan state machine for a phase: Discussion → Research → Planning. Each stage produces GitHub Issues/comments as artifacts, shows a gate summary, and waits for confirmation before advancing.
 </objective>
 
-<execution_context>
-@~/.claude/maxsim/workflows/plan.md
-@~/.claude/maxsim/references/ui-brand.md
-</execution_context>
-
 <context>
-Phase number: $ARGUMENTS (optional — auto-detects next unplanned phase if omitted)
+Arguments: $ARGUMENTS
+
+Phase number is required. Auto-detects next unplanned phase if omitted.
+
+Flags:
+- `--force-research` — Re-run research even if research notes already exist on the GitHub Issue
+- `--skip-verify` — Skip plan verification loop after the planning stage
 
-**Flags:**
-- `--force-research` — Re-run research even if RESEARCH.md exists
-- `--skip-verify` — Skip plan verification loop after planning stage
+GitHub is the sole source of truth. Stage state is read from GitHub Issue labels and comments — no .planning/ files.
 
-Context files are resolved inside the workflow via `maxsim-tools init plan-phase` and per-subagent context assembly.
+Re-entry: If phase is already planned, show status and offer options (view, re-plan, execute).
 </context>
 
 <process>
-Execute the plan workflow from @~/.claude/maxsim/workflows/plan.md end-to-end.
-Preserve all workflow gates (stage detection, discussion, research, planning, gate confirmations, re-entry flow).
+Follow @~/.claude/maxsim/workflows/plan.md end-to-end.
+
+1. Detect current stage from GitHub Issue labels on the phase Issue
+2. Start at earliest incomplete stage
+3. Discussion stage: clarify scope with user via AskUserQuestion
+4. Research stage: spawn Agent for domain/codebase research (skipped if notes exist and no --force-research)
+5. Planning stage: spawn Agent to break phase into task Issues with acceptance criteria
+6. Show gate summary after each stage — wait for confirmation before advancing
 </process>

package/dist/assets/templates/commands/maxsim/progress.md

@@ -1,25 +1,24 @@
 ---
 name: maxsim:progress
-description: Check project progress, milestone status, and route to next action
-allowed-tools:
-  - Read
-  - Bash
-  - Grep
-  - Glob
-  - Task
-  - SlashCommand
+description: Show project status from GitHub Project Board with next-action recommendation
+allowed-tools: [Read, Bash, Grep, Glob]
 ---
-<objective>
-Check project progress, milestone status, and offer milestone completion when all phases are done. Shows GitHub Issues-based progress alongside local ROADMAP.md progress for cross-validation.
 
-Provides situational awareness before continuing work, detects phase gaps, and intelligently routes to the next action.
+<objective>
+Show current project progress from GitHub and recommend the next action. Provides situational awareness, detects gaps, and routes to the appropriate next command.
 </objective>
 
-<execution_context>
-@~/.claude/maxsim/workflows/progress.md
-</execution_context>
+<context>
+GitHub is the sole source of truth. Read the GitHub Project Board, Milestone progress, and open Issues to build the status view — no .planning/ or ROADMAP.md files.
+</context>
 
 <process>
-Execute the progress workflow from @~/.claude/maxsim/workflows/progress.md end-to-end.
-Preserve all routing logic (Routes A through F) and edge case handling.
+Follow @~/.claude/maxsim/workflows/progress.md end-to-end.
+
+1. Read active GitHub Milestone and its completion percentage
+2. Read GitHub Project Board columns to get phase/task states
+3. List open Issues by label (bug, quick, debug, blocked)
+4. Render a status table: phases with planned/executed/verified state
+5. Detect gaps (planned but not executed, executed but not verified)
+6. Recommend next action with the exact command to run
 </process>

package/dist/assets/templates/commands/maxsim/quick.md

@@ -1,41 +1,30 @@
 ---
 name: maxsim:quick
-description: Execute a quick task with MAXSIM guarantees
-argument-hint: "[--full]"
-allowed-tools:
-  - Read
-  - Write
-  - Edit
-  - Glob
-  - Grep
-  - Bash
-  - Task
-  - AskUserQuestion
+description: Quick task - create GitHub Issue and execute in simplified flow
+argument-hint: "[task description]"
+allowed-tools: [Read, Write, Edit, Bash, Grep, Glob, Agent, AskUserQuestion]
 ---
-<objective>
-Execute small, ad-hoc tasks with MAXSIM guarantees (atomic commits, STATE.md tracking).
-
-Quick mode is the same system with a shorter path:
-- Spawns planner (quick mode) + executor(s)
-- Quick tasks are tracked via GitHub Issues (label: "quick"), separate from planned phases
-- Updates STATE.md "Quick Tasks Completed" table (NOT ROADMAP.md)
 
-**Default:** Skips research, plan-checker, verifier. Use when you know exactly what to do.
-
-**`--full` flag:** Enables plan-checking (max 2 iterations) and post-execution verification. Use when you want quality guarantees without full milestone ceremony.
+<objective>
+Execute small, ad-hoc tasks with MaxsimCLI guarantees (atomic commits, GitHub tracking). Skips research, plan-checker, and verifier by default for speed.
 </objective>
 
-<execution_context>
-@~/.claude/maxsim/workflows/quick.md
-</execution_context>
-
 <context>
-$ARGUMENTS
+Arguments: $ARGUMENTS
+
+If $ARGUMENTS is provided, treat it as the task description.
 
-Context files are resolved inside the workflow (`init quick`) and delegated via `<files_to_read>` blocks.
+Quick tasks are tracked as GitHub Issues (label: `quick`) — separate from planned phase Issues. GitHub is the sole source of truth — no STATE.md or .planning/ files.
 </context>
 
 <process>
-Execute the quick workflow from @~/.claude/maxsim/workflows/quick.md end-to-end.
-Preserve all workflow gates (validation, task description, planning, execution, state updates, commits).
+Follow @~/.claude/maxsim/workflows/quick.md end-to-end.
+
+1. Get task description from $ARGUMENTS or via AskUserQuestion
+2. Clarify scope if ambiguous (one focused question)
+3. Create a GitHub Issue labeled `quick` for the task
+4. Spawn a planner Agent (quick mode) to produce a concise implementation plan
+5. Spawn executor Agent(s) to implement the plan
+6. Commit with atomic message referencing the GitHub Issue
+7. Close the GitHub Issue with completion summary
 </process>

package/dist/assets/templates/commands/maxsim/settings.md

@@ -1,37 +1,29 @@
 ---
 name: maxsim:settings
-description: Configure MAXSIM workflow toggles and model profile
-allowed-tools:
-  - Read
-  - Write
-  - Bash
-  - AskUserQuestion
+description: View and modify MaxsimCLI configuration
+allowed-tools: [Read, Write, Edit, Bash, AskUserQuestion]
 ---
 
 <objective>
-Interactive configuration of MAXSIM workflow agents, model profile, and pipeline settings via multi-question prompt. Includes integrated profile management with model assignment details per tier.
-
-Routes to the settings workflow which handles:
-- Config existence ensuring
-- Current settings reading and parsing
-- Interactive prompt (model profile, research, plan_checker, verifier, auto-advance, nyquist, branching)
-- Profile description showing actual model assignments per profile tier
-- Config merging and writing
-- Confirmation display
+Interactively view and modify MaxsimCLI configuration: model profile, pipeline toggles (research, plan-checker, verifier, auto-advance), and branching strategy.
 </objective>
 
-<execution_context>
-@~/.claude/maxsim/workflows/settings.md
-</execution_context>
+<context>
+Configuration is stored in CLAUDE.md (project-level) and `~/.claude/maxsim/config.json` (global). Present current values as pre-selections in each prompt.
+</context>
 
 <process>
-**Follow the settings workflow** from `@~/.claude/maxsim/workflows/settings.md`.
+Follow @~/.claude/maxsim/workflows/settings.md end-to-end.
 
-The workflow handles all logic including:
-1. Config file creation with defaults if missing
-2. Current config reading
-3. Interactive settings presentation with pre-selection
-4. Answer parsing and config merging
-5. File writing
-6. Confirmation display
+1. Read current config from CLAUDE.md and `~/.claude/maxsim/config.json`
+2. Display current settings with descriptions
+3. Use AskUserQuestion to interactively configure:
+   - Model profile (with per-tier model assignment details)
+   - Research stage enabled/disabled
+   - Plan-checker enabled/disabled
+   - Verifier enabled/disabled
+   - Auto-advance between stages
+   - Branching strategy (per-phase / per-task / none)
+4. Merge answers and write updated config
+5. Display confirmation of saved settings
 </process>