@iservu-inc/adf-cli 0.17.5 → 0.18.0

package/.project/chats/current/SESSION-STATUS.md

@@ -1,39 +1,41 @@
 # ADF CLI - Current Session Status
 
-**Date:** 2026-01-13
-**Status:** ✅ Active - v0.14.6 Published
-**Latest Version:** v0.14.6
-**Phase:** Phase 6 Complete - Modernization & Stability
+**Date:** 2026-02-23
+**Status:** ✅ Active - v0.18.0 Publishing
+**Latest Version:** v0.18.0
+**Phase:** Phase 7 Complete - Harness Engineering Protocol
 **Assistant:** Bob (established 2025-10-05)
 
 ---
 
 ## Recent Completion
 
-**Session:** Phase 6 Documentation & Stability Fixes
-**Completed:** 2026-01-13
-**Status:** ✅ v0.14.6 Published to npm
+**Session:** Phase 7 - Harness Engineering Protocol
+**Completed:** 2026-02-23
+**Status:** ✅ v0.18.0 Publishing to npm
 **Branch:** development
 **Major Achievements:**
-- Dynamic AI model discovery for all providers.
-- Migration to `@google/genai` (v1.35.0).
-- Self-healing configuration validation.
-- Agent-Native repository standard (`AGENTS.md` + `.context/`).
-- Removal of deprecated models (Gemini 1.5, GPT-3.5).
+- Harness Engineering Protocol with 7 core data structures
+- Multi-provider session handoff (Anthropic, OpenAI, Google Gemini)
+- Headless/CI automation mode with JSON-driven input
+- 9 new modules, 11 test files, 107 new tests (556 total)
+- Zero regressions, fully opt-in via `--harness` flag
 
 ---
 
 ## Current State
 
 ### Published Version
-- **Package:** `@iservu-inc/adf-cli@0.14.6`
-- **Status:** ✅ Live on npm
+- **Package:** `@iservu-inc/adf-cli@0.18.0`
+- **Status:** ✅ Publishing to npm
 - **Branch:** development
 - **Major Features:**
-  - **Dynamic Model Discovery** - Real-time API fetching for AI models.
-  - **Learning Analytics Dashboard** - 10+ metrics with multi-format export.
-  - **Pattern Decay Algorithm** - Exponential decay for inactive patterns.
-  - **Agent-Native Manifest** - Standardized documentation for AI agents.
+  - **Harness Protocol** - Long-running sessions across context windows
+  - **Cross-Provider Handoff** - Start with Claude, resume with GPT or Gemini
+  - **Headless Mode** - CI/CD pipeline automation
+  - **Structured Events** - JSONL event logging for audit trails
+  - **Feature Manifest** - Passes-only mutation tracking
+  - **Milestone Tracking** - Question block to milestone mapping
 
 ### npm Registry
 ```bash
@@ -41,23 +43,23 @@ npm install -g @iservu-inc/adf-cli@latest
 ```
 
 ### Recent Version History
-- **v0.14.6** (2026-01-12) - Cleanup of deprecated models & help text polish.
-- **v0.14.5** (2026-01-12) - Reconfiguration loop & hang fixes.
-- **v0.14.3** (2026-01-12) - Self-healing config & unified Google SDK.
-- **v0.11.0** (2025-11-04) - Learning Analytics Dashboard (Phase 6.2)
-- **v0.10.0** (2025-10-27) - Pattern Decay Algorithm (Phase 6.1)
+- **v0.18.0** (2026-02-23) - Harness Engineering Protocol (Phase 7)
+- **v0.17.5** (2025-12-23) - A2A protocol, template sync
+- **v0.17.1** (2025-12-23) - Question consolidation
+- **v0.17.0** (2025-12-23) - A2A protocol integration
+- **v0.16.0** (2026-01-21) - Custom artifact import, rules exchange
 
 ---
 
 ## Next Phase Options
 
-### Phase 7: AI-Enhanced Filtering
+### Phase 8: AI-Enhanced Filtering
 - AI-powered relevance scoring (beyond rules)
 - Dynamic question generation
 - Custom follow-ups for project types
 - AI-assisted pattern detection
 
-### Phase 8: Community & Ecosystem
+### Phase 9: Community & Ecosystem
 - Plugin system for custom frameworks
 - Shareable interview templates
 - Community question database
@@ -65,5 +67,5 @@ npm install -g @iservu-inc/adf-cli@latest
 
 ---
 
-**Last Updated:** 2026-01-13 (by Bob)
-**Status:** Phase 6 Complete - Ready for Phase 7 Planning 🚀
+**Last Updated:** 2026-02-23 (by Bob)
+**Status:** Phase 7 Complete - Ready for Phase 8 Planning

package/.project/docs/ROADMAP.md

@@ -1,8 +1,8 @@
 # ADF CLI - Development Roadmap
 
-**Last Updated:** 2026-01-21
-**Current Version:** v0.16.0
-**Status:** Phase 6.2 Complete, Phase 7 Active
+**Last Updated:** 2026-02-23
+**Current Version:** v0.18.0
+**Status:** Phase 7 Complete, Phase 8 Planning
 
 ---
 
@@ -351,14 +351,61 @@ Create the definitive AI-driven requirements gathering tool that transforms vagu
 
 ---
 
+### ✅ Phase 7: Harness Engineering Protocol (v0.18.0)
+**Released:** 2026-02-23
+**Status:** Complete
+
+**Delivered:**
+- Common harness protocol synthesizing patterns from OpenAI Codex, Anthropic Claude Code, and Google Gemini
+- 7 core data structures: HarnessRun, ContextWindow, HandoffPackage, Milestone, Event, FeatureManifest, ProviderCapability
+- Multi-provider support with capability registry (Anthropic 200K/1M, OpenAI 400K, Google 1M)
+- Cross-provider session handoff (start with Claude, resume with GPT or Gemini)
+- Headless/CI automation mode with JSON-driven input
+- 9 new modules, 11 test files, 107 new tests (556 total passing)
+- Zero regressions, fully opt-in via `--harness` flag
+
+**New Commands:**
+- `adf init --harness` / `--headless` / `--run-id <id>`
+- `adf harness start|resume|status|handoff|events|manifest`
+
+**New Modules:**
+- `lib/harness/protocol.js` - Schema classes with validation
+- `lib/harness/event-logger.js` - JSONL append-only event log
+- `lib/harness/run-manager.js` - Run lifecycle management
+- `lib/harness/milestone-tracker.js` - Question block → milestone mapping
+- `lib/harness/feature-manifest.js` - Passes-only feature tracking
+- `lib/harness/context-window-manager.js` - Window lifecycle and handoff
+- `lib/harness/provider-bridge.js` - Multi-provider capability wrapper
+- `lib/harness/headless-adapter.js` - CI/automation JSON input
+- `lib/commands/harness.js` - CLI command handlers
+
+**Modified Modules:**
+- `knowledge-graph.js` - toHandoffFormat/fromHandoffFormat
+- `dynamic-pipeline.js` - serializeForHandoff/deserializeFromHandoff
+- `progress-tracker.js` - Event emission integration
+- `session-manager.js` - getCurrentSession/forkSession
+- `interviewer.js` - setHarness/consumeHandoff/startHeadless
+- `init.js` - Harness flag handling
+- `bin/adf.js` - Harness command group
+
+**Impact:**
+- ✅ Long-running AI sessions across multiple context windows
+- ✅ Provider-agnostic handoff packages
+- ✅ Structured event logging for audit trails
+- ✅ Milestone-based progress tracking
+- ✅ Headless mode for CI/CD pipelines
+- ✅ Full backward compatibility (opt-in only)
+
+---
+
 ## Active Development
 
-### 🔄 Phase 7: AI-Enhanced Filtering & Automation
-**Target Version:** v0.16.0
+### 🔄 Phase 8: AI-Enhanced Filtering & Automation
+**Target Version:** v0.19.0
 **Priority:** Medium
 **Estimated Effort:** 3-4 weeks
 **Status:** Planning
-**Dependencies:** Phase 6 complete
+**Dependencies:** Phase 7 complete
 
 #### Goals
 Use AI to go beyond rule-based filtering for even more intelligent question selection.
@@ -369,67 +416,29 @@ Use AI to go beyond rule-based filtering for even more intelligent question sele
 **Problem:** Rule-based scoring may miss nuanced relevance.
 **Solution:** Use LLM to evaluate question relevance to project.
 
-**Implementation:**
-- Send project context + question to AI
-- Get relevance score (0-100) with reasoning
-- Combine with rule-based score
-- Confidence-based suggestions
-
-**User Benefit:** More accurate relevance determination
-
 ##### 2. Dynamic Question Generation
 **Problem:** Fixed question set may not cover all scenarios.
 **Solution:** Generate questions based on project context.
 
-**Implementation:**
-- Analyze project deeply with AI
-- Generate custom questions for gaps
-- Adaptive interview paths
-- Framework-specific question sets
-
-**User Benefit:** Perfectly tailored interviews
-
 ##### 3. Custom Follow-Ups for Project Types
 **Problem:** Generic follow-ups don't fit all project types.
 **Solution:** Generate context-aware follow-up questions.
 
-**Implementation:**
-- Project-type-specific follow-up templates
-- AI-generated follow-ups based on context
-- Intelligent question chains
-- Deep-dive capabilities
-
-**User Benefit:** More thorough requirements gathering
-
 ##### 4. AI-Assisted Pattern Detection
 **Problem:** Rule-based pattern detection may miss subtle patterns.
 **Solution:** Use AI to find patterns in skip/answer history.
 
-**Implementation:**
-- Semantic similarity clustering
-- Natural language pattern descriptions
-- Anomaly detection
-- Confidence scoring improvements
-
-**User Benefit:** Better pattern recognition
-
 #### Success Criteria
 - AI scoring improves relevance accuracy by >15%
 - Dynamic questions fill gaps in 90% of interviews
 - Custom follow-ups rated higher quality by users
 - AI pattern detection finds 20% more patterns
 
-#### Technical Challenges
-- AI API costs (need caching/batching)
-- Latency (can't slow down interview)
-- Quality control on generated questions
-- Balancing AI vs rules
-
 ---
 
 ## Future Vision
 
-### 🔮 Phase 8: Community & Ecosystem (Long-term)
+### 🔮 Phase 9: Community & Ecosystem (Long-term)
 **Target Version:** v1.0.0
 **Priority:** Low
 **Estimated Effort:** 8-12 weeks
@@ -529,6 +538,7 @@ Build a community-driven ecosystem around ADF CLI with plugins, templates, and s
 - ✅ **Learning Analytics** - User visibility and trust (COMPLETE v0.15.0)
 - ✅ **Export/Import Rules** - Team collaboration (COMPLETE v0.16.0)
 - ✅ **Custom Artifact Import** - Zero-friction onboarding (COMPLETE v0.16.0)
+- ✅ **Harness Engineering Protocol** - Long-running sessions (COMPLETE v0.18.0)
 
 ### Medium Priority (Do Soon)
 - **AI-Powered Relevance** - Improve accuracy
@@ -544,30 +554,25 @@ Build a community-driven ecosystem around ADF CLI with plugins, templates, and s
 
 ## Version Planning
 
-### v0.16.0 - Current Release ✅
-- Custom artifact import (`adf import`)
-- Learning rules export/import
-- Team collaboration features
-- Released: 2026-01-21
+### v0.18.0 - Current Release ✅
+- Harness Engineering Protocol
+- Multi-provider session handoff
+- Headless/CI automation mode
+- 9 new modules, 107 new tests
+- Released: 2026-02-23
 
-### v0.15.1 - Previous Release ✅
-- Interview UI improvements
-- AI response reliability fixes
-- Context review feature
-- Released: 2026-01-13
+### v0.17.5 - Previous Release ✅
+- A2A protocol integration
+- Template sync from AgentDevFramework
+- Question consolidation
+- Released: 2025-12-23
 
-### v0.17.0 - AI Enhancement (Next)
+### v0.19.0 - AI Enhancement (Next)
 - AI-powered relevance scoring
 - Dynamic question generation
 - Custom follow-ups
 - Estimated: 3-4 weeks
 
-### v0.13.0 - AI Pattern Detection
-- AI-assisted pattern detection
-- Semantic clustering
-- Natural language descriptions
-- Estimated: 2 weeks
-
 ### v1.0.0-rc - Pre-Release Polish
 - Bug fixes and optimizations
 - Performance improvements
@@ -665,6 +670,11 @@ Build a community-driven ecosystem around ADF CLI with plugins, templates, and s
 
 ## Changelog
 
+### 2026-02-23
+- Phase 7 complete (v0.18.0) - Harness Engineering Protocol
+- 9 new harness modules with 107 tests (556 total)
+- Multi-provider handoff, headless mode, structured event logging
+
 ### 2026-01-13
 - Phase 6 marked complete (v0.15.0)
 - Pattern Decay system implemented
@@ -681,6 +691,6 @@ Build a community-driven ecosystem around ADF CLI with plugins, templates, and s
 
 ---
 
-**Ready for:** Phase 6 implementation
-**Current Focus:** Advanced learning features (pattern decay, analytics, export/import)
+**Ready for:** Phase 8 planning
+**Current Focus:** AI-enhanced filtering and automation
 **Long-term Vision:** Community-driven ecosystem at v1.0

package/CHANGELOG.md

@@ -5,6 +5,84 @@ All notable changes to `@iservu-inc/adf-cli` will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.18.0] - 2026-02-23
+
+### 🚀 Major Feature - Harness Engineering Protocol
+
+**Minor Release:** Introduces the Harness Engineering Protocol - a comprehensive system for managing long-running AI agent sessions across multiple context windows, providers, and execution modes. Synthesizes patterns from OpenAI Codex, Anthropic Claude Code, and Google Gemini into a unified, provider-agnostic protocol.
+
+#### Added
+
+**Harness Protocol Core (7 Data Structures):**
+- ✨ **HarnessRun** - Top-level session container with lifecycle management (initializing, running, paused, completed, failed)
+- ✨ **ContextWindow** - Bounded work unit with sequence chaining and token tracking
+- ✨ **HandoffPackage** - Provider-agnostic cross-window state transfer (the key innovation)
+- ✨ **Milestone** - Question block checkpoint tracking with acceptance criteria
+- ✨ **Event** - Structured JSONL log entries (12 event types for audit trails)
+- ✨ **FeatureManifest** - Deliverables list with passes-only mutation pattern
+- ✨ **ProviderCapability** - Per-model metadata registry for multi-provider support
+
+**Multi-Provider Support:**
+- ✨ **Provider Bridge** - Wraps AIClient with capability-aware instrumentation
+- ✨ **Anthropic** - Claude Opus 4.6, Sonnet 4.6, Haiku 4.5 (200K standard, 1M beta)
+- ✨ **OpenAI** - GPT-5.2, GPT-5.2 Pro, GPT-5.2 Codex, GPT-5 Mini (up to 400K)
+- ✨ **Google Gemini** - Gemini 3.1 Pro, 3 Pro, 3 Flash, 2.5 Pro (1M context)
+- ✨ **Cross-Provider Handoff** - Start a session with Claude, resume with GPT or Gemini seamlessly
+
+**New CLI Commands:**
+- ✨ **`adf harness start`** - Start a new harness run with workflow and goal configuration
+- ✨ **`adf harness resume`** - Resume a paused run, consuming the latest handoff package
+- ✨ **`adf harness status`** - Show current run status, milestones, and context window info
+- ✨ **`adf harness handoff`** - Generate or consume a handoff package for cross-window continuity
+- ✨ **`adf harness events`** - View structured event log with filtering and statistics
+- ✨ **`adf harness manifest`** - View feature manifest and progress tracking
+
+**New Init Flags:**
+- ✨ **`adf init --harness`** - Enable harness protocol during initialization
+- ✨ **`adf init --headless`** - Run in headless mode for CI/automation
+- ✨ **`adf init --run-id <id>`** - Resume an existing harness run
+
+**Headless/CI Mode:**
+- ✨ **HeadlessAdapter** - Replaces inquirer prompts with JSON-driven input
+- ✨ **Mapped Answers** - Provide answers by question ID
+- ✨ **Ordered Answers** - Provide answers in sequence
+- ✨ **Default Answers** - Dynamic defaults via function callbacks
+
+**Enhanced Existing Modules:**
+- 🔧 **KnowledgeGraph** - New `toHandoffFormat()` / `fromHandoffFormat()` for portable state transfer
+- 🔧 **DynamicPipeline** - New `serializeForHandoff()` / `deserializeFromHandoff()` for pipeline state
+- 🔧 **ProgressTracker** - Event emission integration with optional EventLogger
+- 🔧 **SessionManager** - New `getCurrentSession()` with harness pointer, `forkSession()` for branching
+- 🔧 **Interviewer** - New `setHarness()`, `consumeHandoff()`, `startHeadless()`, SIGINT handoff generation
+
+#### Technical Details
+
+**New Files (9 modules + 11 test files):**
+- `lib/harness/protocol.js` - Schema classes with validation and serialization
+- `lib/harness/event-logger.js` - JSONL append/query engine
+- `lib/harness/run-manager.js` - Run lifecycle with `current-run.json` pointer
+- `lib/harness/milestone-tracker.js` - Question block to milestone mapping
+- `lib/harness/feature-manifest.js` - Feature extraction from markdown outputs
+- `lib/harness/context-window-manager.js` - Window lifecycle and handoff generation
+- `lib/harness/provider-bridge.js` - PROVIDER_CAPABILITIES registry and instrumented calls
+- `lib/harness/headless-adapter.js` - CI/automation JSON input adapter
+- `lib/commands/harness.js` - CLI command handlers
+
+**Testing:**
+- 107 new tests across 11 test files
+- 556 total tests passing across 47 test suites
+- Coverage: 77.6% statements, 62% branches, 73.8% functions, 79.2% lines
+- Zero regressions to existing functionality
+- Integration tests: full lifecycle, backward compatibility, cross-provider handoff
+
+**Backward Compatibility:**
+- All harness features are opt-in via `--harness` flag
+- No `harness/` directory created without explicit activation
+- Existing sessions, progress tracking, and learning system work identically
+- No new required dependencies
+
+---
+
 ## [0.16.0] - 2026-01-21
 
 ### 🚀 Major Feature - Custom Artifact Import & Learning Rules Exchange

package/README.md

@@ -60,6 +60,21 @@ adf init --tool windsurf
 adf init --rapid --tool cursor
 ```
 
+#### Harness Mode (Long-Running Sessions)
+
+Enable the harness protocol for multi-context-window sessions:
+
+```bash
+# Start with harness protocol
+adf init --harness
+
+# Headless mode for CI/automation
+adf init --harness --headless
+
+# Resume an existing harness run
+adf init --run-id run_abc123
+```
+
 ### Configure ADF Settings
 
 Configure ADF settings like AI provider, AI analysis settings, learning system, and more:
@@ -132,6 +147,38 @@ Or update directly:
 npm update -g @iservu-inc/adf-cli
 ```
 
+### Manage Long-Running Sessions (Harness)
+
+The harness protocol enables long-running AI sessions across multiple context windows with cross-provider handoff:
+
+```bash
+# Start a new harness run
+adf harness start --workflow balanced
+
+# Check current run status
+adf harness status
+
+# Resume a paused run
+adf harness resume
+
+# Generate a handoff package for context window transfer
+adf harness handoff
+
+# View structured event log
+adf harness events --stats
+
+# View feature manifest and progress
+adf harness manifest
+```
+
+**Key Features:**
+- **Cross-Provider Handoff** - Start with Claude, resume with GPT or Gemini
+- **Structured Event Logging** - JSONL audit trail with 12 event types
+- **Milestone Tracking** - Question blocks mapped to trackable milestones
+- **Feature Manifest** - Passes-only mutation for deliverable tracking
+- **Headless Mode** - JSON-driven input for CI/CD automation
+- **Provider Capability Registry** - Auto-adapts to provider context limits (200K-1M tokens)
+
 ### Version
 
 Check installed version:
@@ -194,6 +241,7 @@ The AI provider enhances your workflow by:
 - **Intelligent Follow-Up Questions** - Automatically generated based on your responses
 - **Context-Aware Guidance** - Tailored suggestions for your project type
 - **Skip Functionality** - Type "skip" anytime to move forward
+- **Harness Protocol** - Long-running sessions across multiple context windows with cross-provider handoff
 
 ### Changing Providers
 
@@ -369,6 +417,9 @@ your-project/
 │   ├── context.json         # Your workflow configuration
 │   ├── sessions/            # Requirements gathering sessions
 │   ├── learning/            # Learning system data (skip history, patterns, rules)
+│   ├── harness/             # Harness protocol data (when --harness enabled)
+│   │   ├── current-run.json # Active run pointer
+│   │   └── runs/            # Run data, context windows, events
 │   ├── scripts/             # Helper scripts
 │   └── shared/              # Templates, agents, and resources
 │       ├── agents/          # Agent definition files
@@ -498,35 +549,20 @@ When we release updates to the framework:
 
 See [CHANGELOG.md](./CHANGELOG.md) for detailed version history.
 
-**Latest:** v0.14.0 (2026-01-12)
-- **Project Context Synthesis & Extended Tool Support (v0.14.0)** - Minor release
-  - ✨ **Adaptive Project Augmentation** - Synthesizes existing .adf, Agent-Native (BMAD), and OpenSpec project files into unified context.
-  - ✨ **AI-Driven Gap Analysis** - Automatically identifies knowledge gaps and generates targeted questionnaire to fill them.
-  - ✨ **Extended Tool Support** - Native generators for Kiro, Trae, and Codex CLI.
-  - ✨ **Tool Audit Registry** - New `adf tools audit` command to monitor integration features across all 12+ supported tools.
-
-**Previous Releases:**
-- **v0.13.0 (2026-01-03)** - OpenCode Multi-Agent Integration
-  - Complete OpenCode generator rewrite with multi-provider, multi-agent support.
-  - Optimal model assignment and MCP server integration.
-- **v0.12.10 (2025-12-23)** - Critical Bug Fixes (v0.12.10) - Patch release
-  - Fixed EISDIR deployment error for directory-based tools
-  - Improved rate limit error detection for Gemini models
-  - Added free tier model recommendations for Google Gemini
-  - Clear messaging for quota vs availability issues
+**Latest:** v0.18.0 (2026-02-23)
+- **Harness Engineering Protocol (v0.18.0)** - Minor release
+  - ✨ **Long-Running Sessions** - Manage AI sessions across multiple context windows with structured handoff
+  - ✨ **Cross-Provider Handoff** - Start a session with Claude, resume with GPT-5.2 or Gemini 3.1 Pro
+  - ✨ **Headless/CI Mode** - JSON-driven automation for CI/CD pipelines
+  - ✨ **Structured Event Logging** - JSONL audit trail with 12 event types
+  - ✨ **Provider Capability Registry** - Auto-adapts to Anthropic (200K/1M), OpenAI (400K), Google (1M)
+  - ✨ **New Commands** - `adf harness start|resume|status|handoff|events|manifest`
 
 **Previous Releases:**
-- **v0.12.9 (2025-12-23)** - Multi-Tool Expansion & AI Provider Robustness
-  - Added 4 new CLI tools: OpenCode, Gemini CLI, DeepAgent, verified Zed support
-  - Comprehensive AI provider improvements with model verification
-  - Fixed Google Gemini API integration and expanded Anthropic models (3 → 11)
-  - Enhanced help system with detailed documentation
-  - OpenAI/OpenRouter parameter compatibility for GPT-5 and o-series models
-- **v0.12.0 (2025-12-22)** - ANDF Standard Implementation
-  - Complete AGENTS.md standard compliance with YAML frontmatter
-  - .context/ directory support with architecture and glossary generation
-  - Zed Editor and Google Antigravity full generator support
-  - MCP integration and agent configuration
+- **v0.17.5 (2025-12-23)** - A2A protocol integration, template sync, question consolidation
+- **v0.16.0 (2026-01-21)** - Custom artifact import, learning rules exchange
+- **v0.15.0 (2026-01-13)** - Advanced Learning Analytics Dashboard
+- **v0.14.0 (2026-01-12)** - Project Context Synthesis & Extended Tool Support
 
 - **v0.10.0 (2025-10-27)** - Pattern Decay Algorithm
   - Time-based exponential decay for inactive patterns

package/bin/adf.js

@@ -23,6 +23,14 @@ const configCommand = require('../lib/commands/config');
 const guideCommand = require('../lib/commands/guide');
 const toolsCommand = require('../lib/commands/tools');
 const importCommand = require('../lib/commands/import');
+const {
+  harnessStart,
+  harnessResume,
+  harnessStatus,
+  harnessHandoff,
+  harnessEvents,
+  harnessManifest
+} = require('../lib/commands/harness');
 
 const program = new Command();
 
@@ -115,6 +123,9 @@ program
   .option('--balanced', 'Skip questions, use Level 2 (Balanced (OpenSpec))')
   .option('--comprehensive', 'Skip questions, use Level 3 (Comprehensive (Agent-Native))')
   .option('--tool <tool>', 'Specify deployment tool (windsurf, cursor, etc.)')
+  .option('--harness', 'Enable harness engineering protocol for long-running sessions')
+  .option('--headless', 'Run in headless mode (CI/automation, requires --harness)')
+  .option('--run-id <id>', 'Resume a specific harness run by ID')
   .addHelpText('after', `
 ${chalk.cyan.bold('Description:')}
   Start an interactive AI-guided interview to gather project requirements.
@@ -176,6 +187,7 @@ program
   .command('deploy [tool]')
   .description('Deploy framework to specified tool')
   .option('-l, --list', 'List available deployment tools')
+  .option('-r, --recommend', 'Show tool recommendations based on project analysis')
   .addHelpText('after', `
 ${chalk.cyan.bold('Description:')}
   Deploy ADF framework outputs to your AI coding assistant or IDE.
@@ -578,6 +590,48 @@ tools
   .description('Audit all supported tools and features')
   .action(toolsCommand);
 
+// adf harness
+const harness = program.command('harness');
+harness.description('Manage long-running harness sessions across context windows');
+
+harness
+  .command('start')
+  .description('Start a new harness run')
+  .option('-w, --workflow <workflow>', 'Workflow level (rapid, balanced, comprehensive)', 'balanced')
+  .option('--headless', 'Run in headless mode for CI/automation')
+  .option('--goals <goals>', 'Comma-separated list of run goals')
+  .action(harnessStart);
+
+harness
+  .command('resume')
+  .description('Resume a paused harness run')
+  .option('--run-id <id>', 'Specific run ID to resume')
+  .action(harnessResume);
+
+harness
+  .command('status')
+  .description('Show current harness run status')
+  .option('-a, --all', 'Show all runs')
+  .action(harnessStatus);
+
+harness
+  .command('handoff')
+  .description('Generate or consume a handoff package')
+  .option('--consume', 'Consume the latest handoff package')
+  .action(harnessHandoff);
+
+harness
+  .command('events')
+  .description('View harness event log')
+  .option('--stats', 'Show event statistics')
+  .option('--tail <count>', 'Show last N events', '20')
+  .action(harnessEvents);
+
+harness
+  .command('manifest')
+  .description('View feature manifest and progress')
+  .action(harnessManifest);
+
 // Handle unknown commands
 program.on('command:*', () => {
   console.error(chalk.red(`\nInvalid command: ${program.args.join(' ')}`));

package/lib/analysis/dynamic-pipeline.js

@@ -257,6 +257,32 @@ class DynamicPipeline {
   isEnabled() {
     return this.options.enabled;
   }
+
+  /**
+   * Serialize pipeline state for handoff
+   */
+  serializeForHandoff() {
+    return {
+      knowledgeGraph: this.knowledgeGraph.toHandoffFormat(),
+      stats: { ...this.stats },
+      options: { ...this.options }
+    };
+  }
+
+  /**
+   * Restore pipeline state from handoff data
+   */
+  deserializeFromHandoff(data) {
+    if (!data) return;
+
+    if (data.knowledgeGraph) {
+      this.knowledgeGraph.fromHandoffFormat(data.knowledgeGraph);
+    }
+
+    if (data.stats) {
+      this.stats = { ...this.stats, ...data.stats };
+    }
+  }
 }
 
 module.exports = DynamicPipeline;