wogiflow 1.8.3 → 1.8.5

package/.claude/commands/wogi-guided-edit.md

@@ -47,7 +47,7 @@ This command modifies files. It MUST operate within one of these contexts:
 4. **Apply**: Make approved changes
    - **After each file edit**, validate by file type:
      - `.js` files: `node --check <file>` + ESLint
-     - `.ts`/`.tsx` files: `npx tsc --noEmit` + ESLint
+     - `.ts`/`.tsx` files: Run typecheck from `config.scripts.typecheck` (read from `.workflow/config.json`) + ESLint
      - `.json` files: JSON parse check
      - Other types: skip validation
    - Do NOT proceed to next file until current file passes validation

package/.claude/commands/wogi-register.md

@@ -52,7 +52,18 @@ If MCP inspection yields few or no results:
 
 ### Step 3: Build Plugin Entry
 
-From the discovered information, construct:
+From the discovered information, construct each capability entry. For each capability, classify its **mode** and **flowPhases**:
+
+**Mode Classification Guide:**
+| Mode | When to use | Examples |
+|------|-------------|----------|
+| `standalone` | Action works independently, no task context needed | Screenshots, status checks, lookups, notifications, sending messages |
+| `flow-integrated` | Action is part of a development workflow phase | Running tests (validating), design checks (exploring), deploys (completing), linting (validating) |
+| `trigger` | Action feeds events INTO WogiFlow routing | Error alerts → `/wogi-bug`, deployment failures → `/wogi-story`, PR comments → `/wogi-review` |
+
+**flowPhases** (only for `flow-integrated` mode): `exploring`, `coding`, `validating`, `completing`
+
+**outputFormat**: `text` (default), `json` (structured data), `artifact` (file/image output)
 
 ```json
 {
@@ -66,7 +77,10 @@ From the discovered information, construct:
       "description": "What this action does",
       "triggerPhrases": ["send to X", "push to X"],
       "mcpTool": "mcp__server__tool_name or null",
-      "requiresTask": false
+      "requiresTask": false,
+      "mode": "standalone|flow-integrated|trigger",
+      "flowPhases": ["validating"],
+      "outputFormat": "text"
     }
   ],
   "metadata": {
@@ -91,10 +105,11 @@ Source: MCP tools | Web search | Manual
 
 Capabilities discovered (N):
   1. <action>: <description>
-     Triggers: "phrase 1", "phrase 2"
+     Mode: standalone | Triggers: "phrase 1", "phrase 2"
      MCP Tool: mcp__server__tool
 
   2. <action>: <description>
+     Mode: flow-integrated | Phases: validating, completing
      Triggers: "phrase 3"
 
 Trigger phrases (top-level):
@@ -183,3 +198,13 @@ When `config.plugins.autoScanOnSessionStart` is true:
 - Display: `New plugin detected: <name>. Auto-registered with N capabilities.`
 
 Mid-session plugin installs require manual `/wogi-register <name>`.
+
+## Post-Registration: Activate Plugin
+
+After saving the plugin to the registry (Step 5), automatically run `/reload-plugins` to activate the newly registered plugin without requiring a session restart. This ensures the plugin's MCP tools are immediately available.
+
+```
+After registration completes, tell the user:
+"Plugin registered. Running /reload-plugins to activate it immediately."
+Then invoke /reload-plugins.
+```

package/.claude/commands/wogi-review.md

@@ -1473,7 +1473,7 @@ For each issue, follow this cycle:
 │  Run targeted verification:          │
 │    node --check <file>  (syntax)     │
 │    npx eslint <file>    (lint)       │
-│    npx tsc --noEmit     (types)      │
+│    config.scripts.typecheck (types)   │
 │              ↓                       │
 │  If PASS → Mark todo: completed      │
 │  If FAIL → Fix and retry             │

package/.claude/commands/wogi-start.md

@@ -20,7 +20,12 @@ When invoked with a **quoted request** instead of a task ID, assess intent and r
 
 **Plugin Registry Routing**: After command catalog finds no match and `config.plugins.enabled`, check `.workflow/state/plugin-registry.json` for trigger phrase matches (score >= 0.5). Plugin routing has LOWER priority than built-in commands.
 
-**Routing order**: Task ID → Long input gate → Command Catalog → Plugin Registry → Default (`/wogi-story`)
+**Plugin Mode-Aware Routing** (when a plugin match is found):
+- **Standalone** (`mode: 'standalone'`): If `config.plugins.standaloneBypassTask` is true, skip task creation. Clear routing flag, let the AI use the plugin directly. Log the action with `#plugin:<name>` tag by running `node -e "require('./scripts/flow-plugin-registry').logPluginAction({pluginName:'<name>',action:'<action>',mode:'standalone'})"`.
+- **Flow-integrated** (`mode: 'flow-integrated'`): Requires an active task. If no task is in progress, route to `/wogi-story` first. The plugin becomes available during its declared `flowPhases` (injected via phase context prompt).
+- **Trigger** (`mode: 'trigger'`): Route to the appropriate `/wogi-*` command based on the trigger type (error triggers → `/wogi-bug`, deployment triggers → `/wogi-story`, PR triggers → `/wogi-review`). The plugin action feeds INTO WogiFlow routing.
+
+**Routing order**: Task ID → Long input gate → Command Catalog → Plugin Registry (mode-aware) → Default (`/wogi-story`)
 
 ### Command Catalog
 

package/.claude/commands/wogi-statusline-setup.md

@@ -60,6 +60,10 @@ Add or update the `statusLine` section in `~/.claude/settings.json`:
 | `{{task.id}}` | Current WogiFlow task ID (if any) |
 | `{{task.title}}` | Current task title |
 | `{{skill}}` | Currently active skill |
+| `{{worktree.name}}` | Worktree name (if running in --worktree session) |
+| `{{worktree.branch}}` | Worktree branch name |
+| `{{worktree.path}}` | Worktree directory path |
+| `{{worktree.original}}` | Original repo directory |
 
 ### Recommended Formats
 
@@ -78,6 +82,11 @@ Add or update the `statusLine` section in `~/.claude/settings.json`:
 "format": "[{{task.id}}] {{model}} | {{context_window.used_percentage}}% used | {{#if skill}}{{skill}}{{/if}}"
 ```
 
+**With Worktree** (for parallel task execution):
+```json
+"format": "{{#if worktree}}[WT:{{worktree.branch}}] {{/if}}{{#if task}}[{{task.id}}] {{/if}}{{model}} | Ctx: {{context_window.used_percentage}}%"
+```
+
 ## WogiFlow Integration
 
 To have WogiFlow automatically update the status line with current task info, we need to:

package/.claude/docs/commands.md

@@ -39,6 +39,7 @@ When user types these commands, execute the corresponding action immediately.
 | `/wogi-init` | Initialize workflow structure. Create all directories and state files. Use for new projects. |
 | `/wogi-review-fix` | Code review with automatic fixing. Runs full review process then auto-fixes all findings. |
 | `/wogi-peer-review` | Multi-model code review - different AI models review same code for diverse perspectives. |
+| `/wogi-eval` | Multi-judge scoring system — 1 Opus + 2 Sonnet judges score task output across 5 dimensions. |
 | `/wogi-triage` | Interactive walkthrough of review findings from last-review.json. Categorize, dismiss, or create tasks. |
 | `/wogi-onboard` | Analyze existing project with deep temporal analysis, pattern extraction, and state file generation. |
 | `/wogi-rescan` | Re-scan project after external changes. Smart diff: auto-adds new items, auto-removes deleted items, presents conflicts one-by-one. Options: `--dry-run`, `--auto-resolve`, `--category`, `--since`. |

package/.claude/docs/knowledge-base/01-setup-onboarding/README.md

@@ -167,7 +167,7 @@ This analyzes your codebase and populates:
 1. **Review Generated Files**: Check `decisions.md` and `app-map.md` for accuracy
 2. **Run Health Check**: `/wogi-health` to verify setup
 3. **View Tasks**: `/wogi-ready` to see available tasks
-4. **Start Working**: `/wogi-start TASK-001`
+4. **Start Working**: `/wogi-start wf-a1b2c3d4`
 
 ---
 

package/.claude/docs/knowledge-base/01-setup-onboarding/component-indexing.md

@@ -45,34 +45,39 @@ Component indexing solves:
 
 ## Configuration
 
+Component indexing is now configured via the `registries` array in `config.json`. Each registry object defines a type, directories, and ignore patterns:
+
 ```json
 {
-  "componentIndex": {
-    "autoScan": true,                    // Enable auto-scanning
-    "scanOn": [                          // When to scan
-      "sessionStart",                    // On new Claude session
-      "afterTask",                       // After task completion
-      "preCommit"                        // Before git commits (requires hooks)
-    ],
-    "staleAfterMinutes": 60,             // Refresh if older than this
-    "directories": [                     // Where to look
-      "src/components",
-      "src/hooks",
-      "src/services",
-      "src/pages",
-      "src/modules",
-      "app"
-    ],
-    "ignore": [                          // What to skip
-      "*.test.*",
-      "*.spec.*",
-      "*.stories.*",
-      "index.ts",
-      "index.js",
-      "__tests__",
-      "__mocks__"
-    ]
-  }
+  "registries": [
+    {
+      "id": "components",
+      "autoScan": true,                    // Enable auto-scanning
+      "scanOn": [                          // When to scan
+        "sessionStart",                    // On new Claude session
+        "afterTask",                       // After task completion
+        "preCommit"                        // Before git commits (requires hooks)
+      ],
+      "staleAfterMinutes": 60,             // Refresh if older than this
+      "directories": [                     // Where to look
+        "src/components",
+        "src/hooks",
+        "src/services",
+        "src/pages",
+        "src/modules",
+        "app"
+      ],
+      "ignore": [                          // What to skip
+        "*.test.*",
+        "*.spec.*",
+        "*.stories.*",
+        "index.ts",
+        "index.js",
+        "__tests__",
+        "__mocks__"
+      ]
+    }
+  ]
 }
 ```
 
@@ -167,10 +172,13 @@ Based on `scanOn` configuration:
 
 ```json
 {
-  "componentIndex": {
-    "scanOn": ["sessionStart", "afterTask", "preCommit"],
-    "staleAfterMinutes": 60
-  }
+  "registries": [
+    {
+      "id": "components",
+      "scanOn": ["sessionStart", "afterTask", "preCommit"],
+      "staleAfterMinutes": 60
+    }
+  ]
 }
 ```
 
@@ -300,28 +308,34 @@ Check directories are correct:
 ls -la src/components/
 ```
 
-Update config if needed:
+Update the component registry entry in the `registries` array:
 ```json
 {
-  "componentIndex": {
-    "directories": ["components", "lib"]  // Adjust paths
-  }
+  "registries": [
+    {
+      "id": "components",
+      "directories": ["components", "lib"]  // Adjust paths
+    }
+  ]
 }
 ```
 
 ### Too Many Results
 
-Add ignore patterns:
+Add ignore patterns to the component registry entry:
 ```json
 {
-  "componentIndex": {
-    "ignore": [
-      "*.test.*",
-      "*.stories.*",
-      "*.d.ts",
-      "internal/*"
-    ]
-  }
+  "registries": [
+    {
+      "id": "components",
+      "ignore": [
+        "*.test.*",
+        "*.stories.*",
+        "*.d.ts",
+        "internal/*"
+      ]
+    }
+  ]
 }
 ```
 

package/.claude/docs/knowledge-base/01-setup-onboarding/installation.md

@@ -37,7 +37,7 @@ trustedDependencies = ["wogiflow"]
 
 ```
 .workflow/
-├── config.json              # Configuration (200+ options)
+├── config.json              # Configuration (500+ options)
 ├── state/
 │   ├── ready.json          # Task queues
 │   ├── app-map.md          # Component registry
@@ -81,9 +81,11 @@ Check `.workflow/config.json` and adjust settings:
     "strictMode": true,
     "requireStoryForMediumTasks": true
   },
-  "loops": {
-    "enforced": true,
-    "maxRetries": 5
+  "execution": {
+    "loops": {
+      "enforced": true,
+      "maxRetries": 5
+    }
   }
 }
 ```
@@ -125,8 +127,8 @@ The installer creates a balanced config:
 | Setting | Default | Purpose |
 |---------|---------|---------|
 | `enforcement.strictMode` | `true` | Require tasks for implementation |
-| `loops.enforced` | `true` | Enable self-completing loops |
-| `loops.maxRetries` | `5` | Retry failed verifications |
+| `execution.loops.enforced` | `true` | Enable self-completing loops |
+| `execution.loops.maxRetries` | `5` | Retry failed verifications |
 | `durableSteps.enabled` | `true` | Enable crash recovery |
 | `autoLog` | `true` | Auto-log changes |
 | `autoUpdateAppMap` | `true` | Auto-update component registry |

package/.claude/docs/knowledge-base/01-setup-onboarding/onboarding-existing.md

@@ -169,12 +169,12 @@ From known issues:
 {
   "tasks": [
     {
-      "id": "TASK-001",
+      "id": "wf-a1b2c3d4",
       "title": "Login sometimes fails on slow connections",
       "type": "bugfix"
     },
     {
-      "id": "TASK-002",
+      "id": "wf-b2c3d4e5",
       "title": "Dashboard needs performance optimization",
       "type": "refactor"
     }
@@ -242,7 +242,7 @@ Edit files to add:
 ### 5. Start Working
 
 ```bash
-/wogi-start TASK-001
+/wogi-start wf-a1b2c3d4
 ```
 
 ---
@@ -288,12 +288,15 @@ This helps local LLMs understand your project structure.
 
 ### Components Not Detected
 
-Check directory patterns in config:
+Check directory patterns in the `registries` array in config:
 ```json
 {
-  "componentIndex": {
-    "directories": ["src/components", "src/hooks", "src/services"]
-  }
+  "registries": [
+    {
+      "type": "component",
+      "directories": ["src/components", "src/hooks", "src/services"]
+    }
+  ]
 }
 ```
 

package/.claude/docs/knowledge-base/02-task-execution/01-task-planning.md

@@ -45,7 +45,7 @@ User: "Add a login button"
 ┌─────────────────────────────────────────┐
 │ Does task exist in ready.json?          │
 ├─────────────────────────────────────────┤
-│ YES → /wogi-start TASK-XXX              │
+│ YES → /wogi-start wf-XXXXXXXX           │
 │ NO  → Assess size first                 │
 └─────────────────────────────────────────┘
 ```
@@ -102,7 +102,7 @@ Stories define acceptance criteria that drive the execution loop.
 This generates a structured story:
 
 ```markdown
-# [TASK-015] Add user authentication
+# [wf-a1b2c3d4] Add user authentication
 
 ## User Story
 **As a** user
@@ -136,7 +136,7 @@ This generates a structured story:
 1. Story is created with proposed acceptance criteria
 2. User reviews and approves (or modifies)
 3. Task is added to `ready.json` with criteria attached
-4. `/wogi-start TASK-XXX` uses criteria to drive loop
+4. `/wogi-start wf-XXXXXXXX` uses criteria to drive loop
 
 ---
 
@@ -170,13 +170,13 @@ Original story: "Add user authentication"
 
 Decomposed tasks:
 ```
-TASK-015-A: Create AuthService with login/logout methods
-TASK-015-B: Create LoginForm component
-TASK-015-C: Add form validation with error states
-TASK-015-D: Handle loading state during auth
-TASK-015-E: Add error handling for failed auth
-TASK-015-F: Integrate with existing routing
-TASK-015-G: Add session persistence
+wf-a1b2c3d4-01: Create AuthService with login/logout methods
+wf-a1b2c3d4-02: Create LoginForm component
+wf-a1b2c3d4-03: Add form validation with error states
+wf-a1b2c3d4-04: Handle loading state during auth
+wf-a1b2c3d4-05: Add error handling for failed auth
+wf-a1b2c3d4-06: Integrate with existing routing
+wf-a1b2c3d4-07: Add session persistence
 ```
 
 ### Configuration Options
@@ -238,8 +238,8 @@ For complex tasks, analyze multiple implementation approaches before committing.
 | `/wogi-story "title"` | Create story with acceptance criteria |
 | `/wogi-story "title" --deep` | Create story with forced decomposition |
 | `/wogi-ready` | Show available tasks |
-| `/wogi-start TASK-XXX` | Start executing a task |
-| `/wogi-deps TASK-XXX` | Show task dependencies |
+| `/wogi-start wf-XXXXXXXX` | Start executing a task |
+| `/wogi-deps wf-XXXXXXXX` | Show task dependencies |
 
 ---
 

package/.claude/docs/knowledge-base/02-task-execution/02-execution-loop.md

@@ -55,18 +55,20 @@ The execution loop is the core mechanism that ensures task completion. When enab
 
 ```json
 {
-  "loops": {
-    "enabled": true,              // Enable execution loops
-    "enforced": true,             // Cannot exit until complete
+  "execution": {
+    "maxIterations": 20,
+    "maxRetries": 5,
     "blockExitUntilComplete": true,
-    "requireVerification": true,  // Must verify each criterion
-    "blockOnSkip": true,          // Cannot skip without approval
-    "maxRetries": 5,              // Failed verification retries
-    "maxIterations": 20,          // Total loop cycles
-    "commitEvery": 3,             // Checkpoint commits
-    "pauseBetweenScenarios": false,
     "autoInferVerification": true,
-    "fallbackToManual": true
+    "loops": {
+      "enabled": true,
+      "enforced": true,
+      "requireVerification": true,
+      "blockOnSkip": true,
+      "commitEvery": 3,
+      "pauseBetweenScenarios": false,
+      "fallbackToManual": true
+    }
   }
 }
 ```
@@ -106,7 +108,7 @@ Durable sessions persist:
 1. **Task Start**: Creates `durable-session.json` with all steps
 2. **Step Execution**: Updates step status after each action
 3. **Interruption**: Session persists on disk
-4. **Resume**: `/wogi-start TASK-XXX` detects existing session
+4. **Resume**: `/wogi-start wf-XXXXXXXX` detects existing session
 5. **Completion**: Session archived for learning
 
 ### Configuration
@@ -114,10 +116,7 @@ Durable sessions persist:
 ```json
 {
   "durableSteps": {
-    "enabled": true,              // Enable durable sessions
-    "autoResume": true,           // Auto-resume on restart
-    "checkSuspensionsOnStart": true,
-    "defaultMaxAttempts": 5
+    "enabled": true
   }
 }
 ```
@@ -126,7 +125,7 @@ Durable sessions persist:
 
 ```json
 {
-  "taskId": "TASK-015",
+  "taskId": "wf-a1b2c3d4",
   "taskType": "task",
   "startedAt": "2024-01-15T10:30:00Z",
   "steps": [
@@ -155,7 +154,7 @@ Durable sessions persist:
 
 ```bash
 # Normal start detects existing session
-/wogi-start TASK-015
+/wogi-start wf-a1b2c3d4
 # Output: "🔄 Resuming from durable session (3/7 steps completed)"
 
 # Force resume after suspension
@@ -185,10 +184,7 @@ For tasks that require waiting (external reviews, CI runs, etc.), suspend/resume
 ```json
 {
   "suspension": {
-    "enabled": true,
-    "pollIntervalSeconds": 60,    // How often to check poll conditions
-    "maxPollAttempts": 120,       // Max checks before timeout
-    "reminderAfterHours": 24      // Remind about suspended tasks
+    "enabled": true
   }
 }
 ```
@@ -233,13 +229,13 @@ flow suspend --poll \
 
 ## Hybrid Mode
 
-Hybrid mode uses a local LLM for execution while Claude plans, saving 85-95% of tokens.
+Hybrid mode uses a local LLM for execution while Claude plans, with configurable token savings depending on model and task complexity.
 
 ### The Trade-off
 
 | Aspect | Claude Only | Hybrid Mode |
 |--------|-------------|-------------|
-| Token Cost | 100% | 5-15% |
+| Token Cost | 100% | Reduced (varies by model) |
 | Code Quality | Highest | Good (varies by model) |
 | Speed | Fast | Depends on hardware |
 | Context | Full | Limited to prompt |
@@ -256,9 +252,10 @@ Hybrid mode uses a local LLM for execution while Claude plans, saving 85-95% of
 
 ```json
 {
-  "hybrid": {
-    "enabled": true,
-    "executor": {
+  "models": {
+    "hybrid": {
+      "enabled": true,
+      "executor": {
       "type": "local",              // "local" | "cloud"
       "provider": "ollama",         // "ollama" | "lmstudio" | "openai" | etc.
       "providerEndpoint": "http://localhost:11434",
@@ -268,12 +265,13 @@ Hybrid mode uses a local LLM for execution while Claude plans, saving 85-95% of
       "adaptToExecutor": true,      // Adjust plan for model capabilities
       "useAdapterKnowledge": true   // Use learned model behaviors
     },
-    "settings": {
-      "temperature": 0.7,
-      "maxTokens": 4096,
-      "maxRetries": 20,
-      "timeout": 120000,
-      "autoExecute": false          // Require approval before execution
+      "settings": {
+        "temperature": 0.7,
+        "maxTokens": 4096,
+        "maxRetries": 20,
+        "timeout": 120000,
+        "autoExecute": false
+      }
     }
   }
 }
@@ -318,7 +316,7 @@ flow session stats
 # Output:
 # Total sessions: 15
 # Completed: 14
-# Avg tokens saved: 85.3%
+# Avg tokens saved: varies by model
 ```
 
 ---
@@ -331,16 +329,10 @@ Execute independent tasks simultaneously using git worktrees.
 
 ```json
 {
-  "parallel": {
-    "enabled": true,
-    "maxConcurrent": 3,           // Max parallel tasks
-    "autoApprove": false,         // Require approval
-    "requireWorktree": true,      // Isolate in worktrees
-    "showProgress": true,
-    "autoDetect": true,           // Detect parallelizable tasks
-    "autoSuggest": true,          // Suggest when beneficial
-    "autoExecute": false,         // Require approval
-    "minTasksForParallel": 2
+  "parallelExecution": {
+    "parallel": {
+      "enabled": true
+    }
   }
 }
 ```
@@ -357,7 +349,7 @@ Execute independent tasks simultaneously using git worktrees.
 
 ```bash
 # Suggest parallel execution
-/wogi-bulk TASK-001 TASK-002 TASK-003
+/wogi-bulk wf-a1b2c3d4 wf-b2c3d4e5 wf-c3d4e5f6
 
 # Check parallel status
 flow parallel status
@@ -378,7 +370,7 @@ node scripts/flow-task-enforcer.js status
 
 # Output:
 # 📊 Active Loop Session
-# Task: TASK-015
+# Task: wf-a1b2c3d4
 # Iteration: 3
 # Retries: 1
 #

package/.claude/docs/knowledge-base/02-task-execution/03-verification.md

@@ -24,9 +24,8 @@ Auto-inference automatically verifies certain types of criteria without manual c
 
 ```json
 {
-  "loops": {
-    "autoInferVerification": true,    // Enable auto-inference
-    "fallbackToManual": true          // Ask user if can't infer
+  "execution": {
+    "autoInferVerification": true
   }
 }
 ```
@@ -126,7 +125,7 @@ The `smokeTest` gate ensures basic functionality still works after refactoring.
 
 ### Gate Execution
 
-When running `/wogi-done TASK-XXX`:
+When task completion runs (automatically via `/wogi-start`):
 
 ```
 Running quality gates...
@@ -146,7 +145,7 @@ If a gate fails:
 1. Error output is captured
 2. Failure details saved to `.workflow/state/last-failure.json`
 3. Task completion is blocked
-4. Fix issues and retry `/wogi-done`
+4. Fix issues and retry
 
 ```
 Running quality gates...
@@ -181,8 +180,8 @@ Run validation commands after file edits or before commits.
     "afterFileEdit": {
       "enabled": true,
       "commands": {
-        "*.ts": ["npx tsc --noEmit"],
-        "*.tsx": ["npx tsc --noEmit", "npx eslint {file} --fix"],
+        "*.ts": ["npm run type-check"],
+        "*.tsx": ["npm run type-check", "npx eslint {file} --fix"],
         "*.js": ["npx eslint {file} --fix"],
         "*.jsx": ["npx eslint {file} --fix"]
       },