workflow-ai 1.0.44 → 1.0.45

package/README.md

@@ -2,6 +2,8 @@
 
 AI Agent Workflow Coordinator — kanban-based pipeline for AI coding agents.
 
+Система координации AI-агентов через файловую канбан-доску. Автоматически оркестрирует выполнение задач: берёт тикеты из очереди, запускает нужного агента, проверяет результат и генерирует отчёты.
+
 ## Install
 
 ```bash
@@ -22,8 +24,21 @@ workflow run
 
 | Command | Description |
 |---------|-------------|
-| `workflow init` | Initialize `.workflow/` directory with kanban board structure |
-| `workflow run` | Execute the workflow pipeline (process tickets from in-progress to done) |
+| `workflow init [path] [--force]` | Initialize `.workflow/` directory with kanban board structure |
+| `workflow run [options]` | Execute the AI pipeline |
+| `workflow update [path]` | Update global dir and recreate junctions/hardlinks |
+| `workflow eject <skill> [path]` | Eject a skill (copy from global to project) |
+| `workflow list [path]` | List skills with status (shared/ejected/project-only) |
+| `workflow help` | Show help |
+| `workflow version` | Show version |
+
+### Run Options
+
+| Option | Description |
+|--------|-------------|
+| `--plan <plan>` | Plan ID to execute |
+| `--config <path>` | Config file path |
+| `--project <path>` | Project root (default: auto-detect) |
 
 ## Init
 
@@ -32,38 +47,123 @@ The `workflow init` command creates the `.workflow/` directory structure:
 ```
 .workflow/
 ├── config/
-│   └── config.yaml          # Workflow configuration
+│   ├── config.yaml               # Workflow configuration
+│   ├── pipeline.yaml             # Pipeline stages and agents
+│   └── ticket-movement-rules.yaml
 ├── plans/
-│   └── current/             # Current development plans
+│   ├── current/                  # Current development plans
+│   └── archive/                  # Archived plans
 ├── tickets/
-│   ├── backlog/             # Pending tickets
-│   ├── in-progress/         # Active tickets
-│   └── done/                # Completed tickets
-├── reports/                  # Generated reports
-└── logs/                     # Pipeline execution logs
+│   ├── backlog/                  # Awaiting conditions
+│   ├── ready/                    # Ready to execute
+│   ├── in-progress/              # Currently active
+│   ├── blocked/                  # Blocked by dependencies
+│   ├── review/                   # Awaiting review
+│   └── done/                     # Completed
+├── reports/                      # Generated reports
+├── logs/                         # Pipeline execution logs
+├── metrics/                      # Performance metrics
+├── templates/                    # Ticket/plan/report templates
+└── src/
+    ├── skills/                   # Skill instructions (symlinks to global)
+    └── scripts/                  # Automation scripts (hardlinks to global)
 ```
 
-## Run
+## Pipeline
+
+The `workflow run` command executes a multi-stage pipeline:
+
+1. **pick-first-task** — select ticket from ready queue
+2. **check-plan-decomposition** — verify plan is decomposed into tickets
+3. **decompose-plan** — break down plan into tickets (if needed)
+4. **check-conditions** — validate ticket readiness conditions
+5. **move-to-ready** — move tickets from backlog to ready
+6. **pick-next-task** — select next ticket for execution
+7. **move-to-in-progress** — start execution
+8. **check-relevance** — verify ticket is still relevant
+9. **execute-task** — perform the work via AI agent
+10. **move-to-review** — submit for review
+11. **review-result** — validate results against Definition of Done
+12. **increment-task-attempts** — track retry attempts
+13. **move-ticket** — move to done/blocked based on review
+14. **create-report** — generate execution report
+15. **analyze-report / decompose-gaps** — analyze results and iterate
+
+### Supported Agents
+
+| Agent | Description |
+|-------|-------------|
+| `claude-sonnet` | Claude Sonnet — fast model for simple tasks |
+| `claude-opus` | Claude Opus — powerful model for complex tasks |
+| `qwen-code` | Qwen Code — alternative agent |
+| `kilo-code` | Kilo Code — multi-mode agent |
+
+Agents are configured in `configs/pipeline.yaml`.
+
+## Skills
+
+Built-in skills for different task types:
+
+| Skill | Description |
+|-------|-------------|
+| `analyze-report` | Report analysis |
+| `check-relevance` | Ticket relevance verification |
+| `coach` | Skill management and improvement |
+| `create-plan` | Plan creation |
+| `create-report` | Report generation |
+| `decompose-gaps` | Gap decomposition |
+| `decompose-plan` | Plan decomposition into tickets |
+| `deep-research` | Deep research |
+| `execute-task` | Task execution |
+| `review-result` | Result review against DoD |
+
+Skills are stored globally in `~/.workflow/skills/` and symlinked into projects.
+
+Use `workflow eject <skill>` to copy a skill into the project for customization.
+
+## Task Types
+
+| Type | Prefix | Description |
+|------|--------|-------------|
+| `arch` | ARCH | Architecture & planning |
+| `impl` | IMPL | Code implementation |
+| `fix` | FIX | Bug fixes |
+| `review` | REVIEW | Code/documentation review |
+| `docs` | DOCS | Documentation |
+| `admin` | ADMIN | Administrative tasks |
+
+## Configuration
 
-The `workflow run` command executes the pipeline:
+### `configs/config.yaml`
 
-1. Reads tickets from `in-progress/`
-2. Executes each task based on its type (IMPL, FIX, DOCS, REVIEW, etc.)
-3. Moves completed tickets to `done/`
-4. Generates reports in `reports/`
+Main workflow configuration: project info, task types, priorities, statuses, condition types, paths, reporting settings.
 
-## Configuration
+### `configs/pipeline.yaml`
 
-### `.workflow/config/config.yaml`
+Pipeline definition: agents, stages, flow control, goto-logic, retry strategies.
 
-Main workflow configuration:
+### `configs/ticket-movement-rules.yaml`
 
-```yaml
-# Workflow settings
-id_counter: 1
-settings:
-  auto_move_tickets: true
-  generate_reports: true
+Rules for automated ticket movement based on review status.
+
+## Project Structure
+
+```
+workflow-ai/
+├── bin/                    # CLI entry point
+├── src/
+│   ├── cli.mjs             # Command parsing
+│   ├── runner.mjs           # Core pipeline orchestrator
+│   ├── init.mjs             # Project initialization
+│   ├── global-dir.mjs       # Global ~/.workflow/ management
+│   ├── junction-manager.mjs # Symlink/hardlink management
+│   ├── wf-loader.mjs        # Config loader
+│   ├── lib/                 # Utility libraries
+│   └── tests/               # Test suite
+├── configs/                # Configuration files (source)
+├── templates/              # Workflow templates (source)
+├── agent-templates/        # AI agent instruction templates
+└── package.json
 ```
 
 ## Requirements

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "workflow-ai",
-  "version": "1.0.44",
+  "version": "1.0.45",
   "description": "AI Agent Workflow Coordinator — kanban-based pipeline for AI coding agents",
   "type": "module",
   "bin": {

package/src/junction-manager.mjs

@@ -101,6 +101,9 @@ export function createSkillJunctions(globalDir, projectSkillsDir) {
       const skillName = skill.name;
       const targetPath = join(globalSkillsDir, skillName);
       const linkPath = join(projectSkillsDir, skillName);
+      if (existsSync(linkPath) && !isJunction(linkPath)) {
+        continue;
+      }
       createJunction(targetPath, linkPath);
     }
   }