Work Context System

Zero-friction work context tracking for chaotic developers.

Automatically detects what you're working on across all tools—terminal, IDE, browser, email—without any manual logging or activation.

Repository Signposts

• Getting Started: START-HERE.md → orientation + workflow
• Current Tasks: TASK-PHASE-1.md (Phase 1 reference) & run ./scripts/task-status.sh
• Architecture: ARCHITECTURE.md for the process-manager decision + phase plan
• Agents: AGENTS.md (roles, responsibilities, handoff expectations)
• Contributors: CONTRIBUTORS.md (append-only acknowledgements)
• Copilot Instructions: .github/copilot-instructions.md (agent workflow + testing guardrails)
• Memory Bank: memory-bank/ (progress, active context, sessions log)
• Scripts: scripts/README.md (task generator + status helper)

Problem

You jump between projects spontaneously. You have long background processes running. You get interrupted by urgent work. You take unexpected AFK breaks. And sometimes you spend 2 hours on something without realizing it.

Traditional time tracking is friction-heavy. Passive tracking doesn't capture reality. The current system doesn't know what's in your browser tab, where your mental context actually is.

Solution

Watch the process manager. Monitor:

• Active window - What's in focus right now?
• Process hierarchy - What's consuming resources?
• File descriptors - What files/networks are active?
• Working directories - Where are processes running?

Infer context from reality, not assumptions.

Why Process-Manager Approach?

❌ App-specific integrations - Brittle, not portable, maintenance hell ✅ Process-manager awareness - Universal, self-healing, works with any tool

Whether you use Chrome or Firefox, Thunderbird or Gmail, VS Code or Vim—the system detects your actual activity.

Core Idea

$ work-status-now
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  ACTIVE RIGHT NOW
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  📂 Project: stock-v3
  🎯 Work Type: code-review
  🪟 Active: Firefox (github.com PR review)
  ⏱️  Duration: 34 minutes

  🐳 Background: 2 docker containers
  📊 Cognitive Load: Getting busy (3 tasks)

Project Status

Phase: Phase 2 mapping heuristics complete — prepping for daemon integration.

Current Deliverables

• Active window, process tree, and file descriptor detectors emit JSON per spec
• Phase 2 mappers (src/mappers/*.sh) resolve URL/file/process signals → project/work-type candidates
• Inference engine stitches detector + mapper output into project/work-type with confidence scoring + signal trace
• tests/test-detection.sh verifies detectors, mappers, JSON schemas, and enforces <2s detector runtime

Docs to Read Next

• memory-bank/progress.md → timeline + completed vs. pending work
• memory-bank/sessions/2025-11-30.md → latest session summary + handover notes
• src/mappers/ → mapping heuristics for URLs, paths, processes feeding inference

See ARCHITECTURE.md for detailed design decisions.

Quick Start (WIP)

# Start daemon
work-daemon start

# Check current context (real-time)
work-status-now

# View timeline of what you worked on
work-timeline

# Analyze your work patterns
work-analyze

Could Be...

• Internal tool for your workflow
• External product (CLI dev-tool)
• Integration point (Slack status, calendar blocking, etc)

Development

Quick Setup with Codespaces

Click the badge above to launch a fully configured development environment in GitHub Codespaces with Node.js 20, bash, jq, and all required tools pre-installed.

Local Development

Requirements:

• Bash 4.0+
• jq
• xdotool, wmctrl (for window detection)
• Node.js 20+ (optional, for web UI)

See ARCHITECTURE.md for implementation plan.

Phase 1: Core detection (active window, process tree, FDs)
Phase 2: Project mapping & inference
Phase 3: Daemon integration
Phase 4: Output & UI