bmad-auto-copilot 1.1.2 → 1.2.0

package/README.md

@@ -30,7 +30,7 @@ The automation loop continuously:
 3. Auto-commits code after successful code reviews
 4. Repeats until all stories reach `done`
 
-Each session uses **Claude Opus 4.6** by default (configurable).
+Each session uses **GPT-5.3-Codex** with **xhigh reasoning effort** by default (configurable).
 
 ## Prerequisites
 
@@ -82,6 +82,7 @@ Installs to: `.scripts/bmad-auto/copilot/`
 npx bmad-auto-copilot run                    # Run with defaults
 npx bmad-auto-copilot run --max 100          # Limit to 100 iterations
 npx bmad-auto-copilot run --model gpt-4.1    # Use different model
+npx bmad-auto-copilot run --reasoning-effort high # Use different reasoning effort
 npx bmad-auto-copilot run --verbose --dry    # Dry run with debug output
 ```
 
@@ -124,7 +125,8 @@ After installation, you can run the scripts directly:
 
 | Variable | Default | Description |
 |----------|---------|-------------|
-| `COPILOT_MODEL` | `claude-opus-4.6` | AI model for all sessions |
+| `COPILOT_MODEL` | `gpt-5.3-codex` | AI model for all sessions |
+| `COPILOT_REASONING_EFFORT` | `xhigh` | Reasoning effort for all sessions |
 | `ITERATION_DELAY` | `10` | Seconds between iterations |
 
 ### bmad-auto-config.yaml
@@ -142,8 +144,8 @@ implementation_artifacts: "_bmad-output/implementation-artifacts"
 Each step creates a **completely isolated Copilot CLI session**:
 
 ```
-copilot --agent sm  --model claude-opus-4.6 -p <prompt> --allow-all --no-ask-user
-copilot --agent dev --model claude-opus-4.6 -p <prompt> --allow-all --no-ask-user
+copilot --model gpt-5.3-codex --reasoning-effort xhigh -p <prompt> --allow-all --no-ask-user
+copilot --model gpt-5.3-codex --reasoning-effort xhigh -p <prompt> --allow-all --no-ask-user
 ```
 
 This ensures:

package/bin/cli.js

@@ -30,7 +30,8 @@ COMMANDS:
   run [path]          Run the automation loop (shortcut for bmad-loop.sh)
 
 OPTIONS:
-  --model <model>     AI model (default: claude-opus-4.6)
+  --model <model>     AI model (default: gpt-5.3-codex)
+  --reasoning-effort  Reasoning effort (default: xhigh)
   --max <n>           Max iterations (default: 250)
   --output <folder>   BMAD output folder (default: auto-detect from config)
   --force             Overwrite existing installation
@@ -39,7 +40,8 @@ OPTIONS:
 EXAMPLES:
   npx bmad-auto-copilot install
   npx bmad-auto-copilot install ~/projects/my-app
-  npx bmad-auto-copilot install --model claude-sonnet-4.5
+  npx bmad-auto-copilot install --model gpt-5.3-codex
+  npx bmad-auto-copilot run --reasoning-effort xhigh
   npx bmad-auto-copilot run --max 100
   npx bmad-auto-copilot status
   npx bmad-auto-copilot uninstall
@@ -54,7 +56,8 @@ function parseArgs(args) {
   const parsed = {
     command: null,
     projectPath: null,
-    model: 'claude-opus-4.6',
+    model: 'gpt-5.3-codex',
+    reasoningEffort: 'xhigh',
     maxIterations: 250,
     outputFolder: null,
     force: false,
@@ -82,6 +85,9 @@ function parseArgs(args) {
       case '--model':
         parsed.model = args[++i];
         break;
+      case '--reasoning-effort':
+        parsed.reasoningEffort = args[++i];
+        break;
       case '--max':
         parsed.maxIterations = parseInt(args[++i], 10);
         break;

package/lib/run.js

@@ -3,7 +3,7 @@
 // ============================================================
 
 const path = require('path');
-const { execSync, spawn } = require('child_process');
+const { spawn } = require('child_process');
 const {
   resolveProject,
   getInstallPath,
@@ -40,8 +40,9 @@ async function run(args) {
         String(args.maxIterations),
         ...(args.verbose ? ['-Verbose'] : []),
         ...(args.dryRun ? ['-DryRun'] : []),
-        ...(args.model !== 'claude-opus-4.6'
-          ? ['-Model', args.model]
+        ...(args.model ? ['-Model', args.model] : []),
+        ...(args.reasoningEffort
+          ? ['-ReasoningEffort', args.reasoningEffort]
           : []),
       ]
     );
@@ -63,6 +64,9 @@ async function run(args) {
   if (args.model) {
     env.COPILOT_MODEL = args.model;
   }
+  if (args.reasoningEffort) {
+    env.COPILOT_REASONING_EFFORT = args.reasoningEffort;
+  }
 
   // Spawn the loop script as a child process with inherited stdio
   const child = spawn(scriptArgs[0], scriptArgs[1], {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bmad-auto-copilot",
-  "version": "1.1.2",
+  "version": "1.2.0",
   "description": "Automated BMAD story development loop using GitHub Copilot CLI — install into any BMAD project",
   "keywords": [
     "bmad",

package/templates/bmad-loop.ps1

@@ -7,11 +7,13 @@
     Each step runs as a NEW Copilot CLI session (fresh context window).
     Continues until all stories reach 'done' state in sprint-status.yaml.
 
-    Model: Claude Opus 4.6 (configurable via -Model or COPILOT_MODEL env var)
+    Model: GPT-5.3-Codex with xhigh reasoning (configurable via params or env vars)
 .PARAMETER MaxIterations
     Safety limit for maximum loop iterations (default: 250)
 .PARAMETER Model
-    AI model to use (default: claude-opus-4.6)
+    AI model to use (default: gpt-5.3-codex)
+.PARAMETER ReasoningEffort
+    Reasoning effort to use (default: xhigh)
 .PARAMETER Verbose
     Enable detailed debug logging
 .PARAMETER DryRun
@@ -31,6 +33,7 @@
 param(
     [int]$MaxIterations = 250,
     [string]$Model = "",
+    [string]$ReasoningEffort = "",
     [switch]$Verbose,
     [switch]$DryRun
 )
@@ -51,7 +54,15 @@ if (-not $Model) {
     $Model = $env:COPILOT_MODEL
 }
 if (-not $Model) {
-    $Model = "claude-opus-4.6"
+    $Model = "gpt-5.3-codex"
+}
+
+# Reasoning effort: prefer parameter > env var > default
+if (-not $ReasoningEffort) {
+    $ReasoningEffort = $env:COPILOT_REASONING_EFFORT
+}
+if (-not $ReasoningEffort) {
+    $ReasoningEffort = "xhigh"
 }
 
 # Delay between iterations (seconds)
@@ -181,6 +192,10 @@ function Get-NextAction {
         Write-Log "[NEXT] Found story in REVIEW state -> code-review (Dev agent)" "Green"
         return "code-review"
     }
+    if ($inProgressCount -gt 0) {
+        Write-Log "[NEXT] Found story in IN-PROGRESS state -> dev-story resume (Dev agent)" "Green"
+        return "dev-story"
+    }
     if ($readyCount -gt 0) {
         Write-Log "[NEXT] Found story in READY-FOR-DEV state -> dev-story (Dev agent)" "Green"
         return "dev-story"
@@ -194,10 +209,6 @@ function Get-NextAction {
         return "complete"
     }
 
-    if ($inProgressCount -gt 0) {
-        Write-Log "[WAIT] Found $inProgressCount story(ies) in-progress" "Yellow"
-    }
-
     return "wait"
 }
 
@@ -220,10 +231,10 @@ function Invoke-CopilotSession {
     $prompt = Get-Content $PromptFile -Raw
     $prompt = $prompt -replace '\{TIMESTAMP\}', (Get-Date -Format "yyyy-MM-dd HH:mm:ss")
 
-    Write-Log "[SESSION] New Copilot CLI session -> Agent: $Agent | Model: $Model | Action: $ActionLabel" "Magenta"
+    Write-Log "[SESSION] New Copilot CLI session -> Agent: $Agent | Model: $Model | Reasoning: $ReasoningEffort | Action: $ActionLabel" "Magenta"
 
     if ($DryRun) {
-        Write-Log "[DRY RUN] Would execute: copilot --agent $Agent --model $Model -p '...' --allow-all --no-ask-user" "Yellow"
+        Write-Log "[DRY RUN] Would execute: copilot --model $Model --reasoning-effort $ReasoningEffort -p '...' --allow-all --no-ask-user" "Yellow"
         return $true
     }
 
@@ -235,6 +246,7 @@ function Invoke-CopilotSession {
         # The prompt itself contains all workflow instructions.
         copilot `
             --model $Model `
+            --reasoning-effort $ReasoningEffort `
             --allow-all `
             --no-ask-user `
             -p $prompt
@@ -343,6 +355,7 @@ Write-Host ""
 Write-Log "[START] BMAD Auto Loop Started" "Green"
 Write-Log "Max iterations: $MaxIterations" "Gray"
 Write-Log "Model: $Model" "Gray"
+Write-Log "Reasoning effort: $ReasoningEffort" "Gray"
 Write-Log "Project root: $ProjectRoot" "Gray"
 Write-Log "Sprint status: $SprintStatusPath" "Gray"
 if ($DryRun) {

package/templates/bmad-loop.sh

@@ -22,7 +22,7 @@
 #   - BMAD installed with .github/agents/ directory (sm, dev agents)
 #   - Sprint planning completed (sprint-status.yaml exists)
 #
-# MODEL: Claude Opus 4.6 (configurable via COPILOT_MODEL env var)
+# MODEL: GPT-5.3-Codex with xhigh reasoning (configurable via env vars)
 #
 # ============================================================
 
@@ -52,8 +52,9 @@ CONFIG_PATH="$SCRIPT_DIR/bmad-auto-config.yaml"
 PROGRESS_LOG="$SCRIPT_DIR/bmad-progress.log"
 PROMPT_DIR="$SCRIPT_DIR/prompts"
 
-# Model configuration - Claude Opus 4.6
-COPILOT_MODEL="${COPILOT_MODEL:-claude-opus-4.6}"
+# Model configuration - GPT-5.3-Codex + xhigh reasoning effort
+COPILOT_MODEL="${COPILOT_MODEL:-gpt-5.3-codex}"
+COPILOT_REASONING_EFFORT="${COPILOT_REASONING_EFFORT:-xhigh}"
 
 # Delay between iterations (seconds)
 ITERATION_DELAY="${ITERATION_DELAY:-10}"
@@ -231,10 +232,10 @@ invoke_copilot() {
     local prompt=$(cat "$prompt_file")
     prompt="${prompt//\{TIMESTAMP\}/$(date "+%Y-%m-%d %H:%M:%S")}"
 
-    log "[SESSION] New Copilot CLI session → Model: $COPILOT_MODEL | Action: $action_label" "magenta"
+    log "[SESSION] New Copilot CLI session → Model: $COPILOT_MODEL | Reasoning: $COPILOT_REASONING_EFFORT | Action: $action_label" "magenta"
 
     if $DRY_RUN; then
-        log "[DRY RUN] Would execute: copilot --model $COPILOT_MODEL -p '...' --allow-all --no-ask-user" "yellow"
+        log "[DRY RUN] Would execute: copilot --model $COPILOT_MODEL --reasoning-effort $COPILOT_REASONING_EFFORT -p '...' --allow-all --no-ask-user" "yellow"
         return 0
     fi
 
@@ -243,6 +244,7 @@ invoke_copilot() {
     # The prompt itself contains all workflow instructions.
     if copilot \
         --model "$COPILOT_MODEL" \
+        --reasoning-effort "$COPILOT_REASONING_EFFORT" \
         --allow-all \
         --no-ask-user \
         -p "$prompt" \
@@ -333,6 +335,7 @@ log "" "white"
 log "[START] BMAD Auto Loop Started" "green"
 log "Max iterations: $MAX_ITERATIONS" "gray"
 log "Model: $COPILOT_MODEL" "gray"
+log "Reasoning effort: $COPILOT_REASONING_EFFORT" "gray"
 log "Project root: $PROJECT_ROOT" "gray"
 log "Sprint status: $SPRINT_STATUS_PATH" "gray"
 if $DRY_RUN; then

package/templates/prompts/code-review.md

@@ -6,9 +6,12 @@ Timestamp: {TIMESTAMP}
 
 ## Step 1: Load Configuration and Workflow
 Read the BMAD config from: `_bmad/bmm/config.yaml` — resolve all variables ({project-root}, {output_folder}, etc.)
-Read the workflow definition from: `_bmad/bmm/workflows/4-implementation/code-review/workflow.yaml`
-Read the workflow instructions from: `_bmad/bmm/workflows/4-implementation/code-review/instructions.xml`
-Read the workflow checklist from: `_bmad/bmm/workflows/4-implementation/code-review/checklist.md`
+Read the workflow definition from: `_bmad/bmm/workflows/4-implementation/bmad-code-review/workflow.md`
+Read and follow these review step files:
+- `_bmad/bmm/workflows/4-implementation/bmad-code-review/steps/step-01-gather-context.md`
+- `_bmad/bmm/workflows/4-implementation/bmad-code-review/steps/step-02-review.md`
+- `_bmad/bmm/workflows/4-implementation/bmad-code-review/steps/step-03-triage.md`
+- `_bmad/bmm/workflows/4-implementation/bmad-code-review/steps/step-04-present.md`
 
 ## Step 2: Find Next Story
 Read `_bmad-output/implementation-artifacts/sprint-status.yaml` and find the FIRST story in `review` state (scan top to bottom, exclude retrospectives).
@@ -24,8 +27,8 @@ Perform a thorough ADVERSARIAL code review following the workflow instructions:
 - Do NOT review files in `_bmad/`, `_bmad-output/`, `.cursor/`, `.windsurf/`, or `.claude/` directories — only review application source code
 - Verify EVERY task marked [x] is actually implemented
 - Verify EVERY Acceptance Criterion is actually satisfied
-- Find 3-10 specific problems: code quality, test coverage, architecture compliance, security, performance
-- **AUTOMATICALLY FIX all HIGH and MEDIUM issues** (this is option [1] from instructions.xml Step 4 — always choose this)
+- Run the review layers and triage to identify specific problems: code quality, test coverage, architecture compliance, security, performance
+- **AUTOMATICALLY FIX all HIGH and MEDIUM issues** as an automation override after triage
 - Run all tests to verify fixes pass
 - Update the story File List and Dev Agent Record with any fixes applied
 - NEVER accept "looks good" without finding real issues
@@ -39,7 +42,7 @@ Based on review outcome:
 ## Rules
 - Use best judgment for ALL decisions — never ask the user
 - Execute ALL steps without pausing — this is YOLO mode (no confirmation between steps)
-- When instructions.xml has `<ask>` tags, always choose "Fix them automatically" (option 1)
+- When workflow step files contain `<ask>` tags or checkpoint confirmations, use best judgment and continue automatically
 - Document issues you cannot auto-fix in the story file
 - All tests must pass after any fixes
 - Follow the workflow checklist to verify completeness

package/templates/prompts/create-story.md

@@ -6,16 +6,16 @@ Timestamp: {TIMESTAMP}
 
 ## Step 1: Load Configuration and Workflow
 Read the BMAD config from: `_bmad/bmm/config.yaml` — resolve all variables ({project-root}, {output_folder}, etc.)
-Read the workflow definition from: `_bmad/bmm/workflows/4-implementation/create-story/workflow.yaml`
-Read the workflow instructions from: `_bmad/bmm/workflows/4-implementation/create-story/instructions.xml`
-Read the workflow template from: `_bmad/bmm/workflows/4-implementation/create-story/template.md`
-Read the workflow checklist from: `_bmad/bmm/workflows/4-implementation/create-story/checklist.md`
+Read the workflow definition from: `_bmad/bmm/workflows/4-implementation/bmad-create-story/workflow.md`
+Read the input discovery protocol from: `_bmad/bmm/workflows/4-implementation/bmad-create-story/discover-inputs.md`
+Read the workflow template from: `_bmad/bmm/workflows/4-implementation/bmad-create-story/template.md`
+Read the workflow checklist from: `_bmad/bmm/workflows/4-implementation/bmad-create-story/checklist.md`
 
 ## Step 2: Find Next Story
 Read `_bmad-output/implementation-artifacts/sprint-status.yaml` and find the FIRST story in `backlog` state (scan top to bottom, exclude retrospectives).
 
 ## Step 3: Load and Analyze Context (Exhaustive)
-Follow the instructions.xml discover_inputs protocol:
+Follow the workflow.md analysis flow and discover-inputs protocol:
 - Read the corresponding epic file from `_bmad-output/planning-artifacts/`
 - Read PRD, architecture docs, and any referenced project context artifacts
 - Analyze previous story files for context continuity (patterns, conventions, decisions)
@@ -32,5 +32,5 @@ Update `_bmad-output/implementation-artifacts/sprint-status.yaml` — change the
 - Do NOT implement the story code — only create the story file
 - Follow the workflow template precisely
 - Execute ALL steps without pausing — this is YOLO mode (no confirmation between steps)
-- When instructions.xml has `<ask>` tags, use best judgment instead of asking
+- When workflow files contain `<ask>` tags or checkpoint confirmations, use best judgment and continue automatically
 - If blocking issues arise, document them in the story file and continue

package/templates/prompts/dev-story.md

@@ -6,10 +6,8 @@ Timestamp: {TIMESTAMP}
 
 ## Step 1: Load Configuration and Workflow
 Read the BMAD config from: `_bmad/bmm/config.yaml` — resolve all variables ({project-root}, {output_folder}, etc.)
-Read the workflow definition from: `_bmad/bmm/workflows/4-implementation/dev-story/workflow.yaml`
-Read the workflow instructions from: `_bmad/bmm/workflows/4-implementation/dev-story/instructions.xml`
-Read the workflow template (if exists) from: `_bmad/bmm/workflows/4-implementation/dev-story/`
-Read the workflow checklist from: `_bmad/bmm/workflows/4-implementation/dev-story/checklist.md`
+Read the workflow definition from: `_bmad/bmm/workflows/4-implementation/bmad-dev-story/workflow.md`
+Read the workflow checklist from: `_bmad/bmm/workflows/4-implementation/bmad-dev-story/checklist.md`
 
 ## Step 2: Find Next Story
 Read `_bmad-output/implementation-artifacts/sprint-status.yaml` and find the FIRST story in `ready-for-dev` or `in-progress` state (scan top to bottom, exclude retrospectives).
@@ -25,7 +23,7 @@ Update `_bmad-output/implementation-artifacts/sprint-status.yaml` — change the
 
 ## Step 5: Implement
 Follow the workflow instructions to implement ALL tasks and subtasks defined in the story:
-- Follow the red-green-refactor TDD cycle as specified in instructions.xml
+- Follow the red-green-refactor TDD cycle as specified in workflow.md
 - Write failing tests first, then minimal code to pass, then refactor
 - Write comprehensive unit tests — tests must pass 100%
 - Follow ALL acceptance criteria precisely
@@ -43,7 +41,7 @@ After ALL tasks are complete and ALL tests pass:
 ## Rules
 - Use best judgment for ALL decisions — never ask the user
 - Execute ALL steps without pausing — this is YOLO mode (no confirmation between steps)
-- When instructions.xml has `<ask>` tags, use best judgment instead of asking
+- When workflow files contain `<ask>` tags or checkpoint confirmations, use best judgment and continue automatically
 - Follow the workflow checklist to verify completeness
 - Ensure all existing tests continue to pass
 - NEVER mark a task [x] unless it is truly complete with passing tests