feat(workflow): add project-level state management with intelligent initialization

Implement comprehensive project.json system for tracking project state, completed features, and architecture overview.

## New Features

### 1. /workflow:init Command (NEW)
- Independent project initialization command using cli-explore-agent
- Deep Scan mode analyzes: technology stack, architecture, key components, metrics
- Generates project.json with comprehensive overview field
- Supports --regenerate flag to update analysis while preserving features
- One-time initialization with intelligent project cognition

### 2. Enhanced session:start
- Added Step 0: Check project.json existence before session creation
- Auto-calls /workflow:init if project.json missing
- Separated project-level vs session-level initialization responsibilities

### 3. Enhanced session:complete
- Added Phase 3: Update Project Feature Registry
- Extracts feature metadata from IMPL_PLAN.md
- Records completed sessions as features in project.json
- Includes: title, description, tags, timeline, traceability (archive_path, commit_hash)
- Auto-generates feature IDs and tags from implementation plans

### 4. Enhanced status Command
- Added --project flag for project overview mode
- Displays: technology stack, architecture patterns, key components, metrics
- Shows completed features with full traceability
- Provides query commands for feature exploration

### 5. Enhanced context-gather
- Integrated project.json reading in context-search-agent workflow
- Phase 1: Load project.json overview as foundational context
- Phase 3: Populate context-package.json from project.json
- Prioritizes project.json context for architecture and tech stack
- Avoids redundant project analysis on every planning session

## Data Flow
project.json (init) → context-gather → context-package.json → task-generation
session-complete → project.json (features registry)

## Benefits
- Single source of truth for project state
- Efficient context reuse across sessions
- Complete feature traceability and history
- Consistent architectural baseline for all workflows

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

.claude/commands/workflow/tools/context-gather.md

@@ -71,9 +71,10 @@ You are executing as context-search-agent (.claude/agents/context-search-agent.m
 Execute complete context-search-agent workflow for implementation planning:
 
 ### Phase 1: Initialization & Pre-Analysis
-1. **Detection**: Check for existing context-package (early exit if valid)
-2. **Foundation**: Initialize code-index, get project structure, load docs
-3. **Analysis**: Extract keywords, determine scope, classify complexity
+1. **Project State Loading**: Read and parse `.workflow/project.json`. Use its `overview` section as the foundational `project_context`. This is your primary source for architecture, tech stack, and key components. If file doesn't exist, proceed with fresh analysis.
+2. **Detection**: Check for existing context-package (early exit if valid)
+3. **Foundation**: Initialize code-index, get project structure, load docs
+4. **Analysis**: Extract keywords, determine scope, classify complexity based on task description and project state
 
 ### Phase 2: Multi-Source Context Discovery
 Execute all 4 discovery tracks:
@@ -84,16 +85,17 @@ Execute all 4 discovery tracks:
 
 ### Phase 3: Synthesis, Assessment & Packaging
 1. Apply relevance scoring and build dependency graph
-2. Synthesize 4-source data (archive > docs > code > web)
-3. Integrate brainstorm artifacts (if .brainstorming/ exists, read content)
-4. Perform conflict detection with risk assessment
-5. **Inject historical conflicts** from archive analysis into conflict_detection
-6. Generate and validate context-package.json
+2. **Synthesize 4-source data**: Merge findings from all sources (archive > docs > code > web). **Prioritize the context from `project.json`** for architecture and tech stack unless code analysis reveals it's outdated.
+3. **Populate `project_context`**: Directly use the `overview` from `project.json` to fill the `project_context` section of the output `context-package.json`. Include technology_stack, architecture, key_components, and entry_points.
+4. Integrate brainstorm artifacts (if .brainstorming/ exists, read content)
+5. Perform conflict detection with risk assessment
+6. **Inject historical conflicts** from archive analysis into conflict_detection
+7. Generate and validate context-package.json
 
 ## Output Requirements
 Complete context-package.json with:
 - **metadata**: task_description, keywords, complexity, tech_stack, session_id
-- **project_context**: architecture_patterns, coding_conventions, tech_stack
+- **project_context**: architecture_patterns, coding_conventions, tech_stack (sourced from `project.json` overview)
 - **assets**: {documentation[], source_code[], config[], tests[]} with relevance scores
 - **dependencies**: {internal[], external[]} with dependency graph
 - **brainstorm_artifacts**: {guidance_specification, role_analyses[], synthesis_output} with content
@@ -139,7 +141,7 @@ Refer to `context-search-agent.md` Phase 3.7 for complete `context-package.json`
 
 **Key Sections**:
 - **metadata**: Session info, keywords, complexity, tech stack
-- **project_context**: Architecture patterns, conventions, tech stack
+- **project_context**: Architecture patterns, conventions, tech stack (populated from `project.json` overview)
 - **assets**: Categorized files with relevance scores (documentation, source_code, config, tests)
 - **dependencies**: Internal and external dependency graphs
 - **brainstorm_artifacts**: Brainstorm documents with full content (if exists)
@@ -277,6 +279,7 @@ if (historicalConflicts.length > 0 && currentRisk === "low") {
 ## Notes
 
 - **Detection-first**: Always check for existing package before invoking agent
+- **Project.json integration**: Agent reads `.workflow/project.json` as primary source for project context, avoiding redundant analysis
 - **Agent autonomy**: Agent handles all discovery logic per `.claude/agents/context-search-agent.md`
 - **No redundancy**: This command is a thin orchestrator, all logic in agent
 - **Plan-specific**: Use this for implementation planning; brainstorm mode uses direct agent call