Skip to content
Back to skills

2487 Project Context 3c779b49

ASecurity

*Last Updated: 2025-12-13*

  • 9 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 11, 2026
developmentpythontesting

Works with

  • claude code
  • terminal
  • mcp

Security analysis

A100/100

Scanned October 11, 2026

npx -y skills add tools-only/X-Skills --skill 2487-project-context_3c779b49 --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of 2487 Project Context 3c779b49?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for 2487 Project Context 3c779b49
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/tools-only-2487-project-context-3c779b49/badge)](https://www.skillsdirectory.com/skills/tools-only-2487-project-context-3c779b49)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

SKILL.md
# Project Context - Claude Code Usage Monitor

*Last Updated: 2025-12-13*

## Current Status

| Item | Status |
|------|--------|
| Phase | Initial Development |
| Current Task | Human-centered development environment complete |
| Environment | Development |
| Tests | Not yet configured |

## Recent Changes

- [2025-12-13] Human-Centered Development Integration
  - Created `01-human-centered-development.md` philosophy doc
  - Renumbered all best practices files (claude-code-config is now 02)
  - Updated configuration guide with mandatory Human Advocate requirements
  - Created UX Designer profile, instructions, and subagent
  - Created UI Designer profile, instructions, and subagent
  - Updated Service Designer to route to UX Designer
  - Added full design trio (SD, UX, UI) to this project

- [2025-12-13] Set up AI-enabling development environment
  - Created `.claude/` directory structure (agents, reference, commands)
  - Updated CLAUDE.md with three-tier architecture
  - Created project context file
  - Added agent definitions
  - Created context-refs quick reference cards

## Environment Status

| Component | Status | Notes |
|-----------|--------|-------|
| Python | Ready | 3.9+ required |
| uv | Required | Package manager |
| Rich | Ready | Terminal UI library |
| pytest | Required | Install with dev dependencies |

## Active Development Areas

1. **Core Monitoring** (`src/claude_monitor/core/`)
   - Token usage calculations
   - ML-based predictions
   - Plan detection

2. **Data Handling** (`src/claude_monitor/data/`)
   - Claude usage data parsing
   - Agent data analysis

3. **UI Components** (`src/claude_monitor/ui/`)
   - Rich terminal interface
   - Layout management
   - Agent display panels

## Known Issues

*None currently tracked*

## Architecture Decisions

| Decision | Rationale | Date |
|----------|-----------|------|
| Human-centered development | Human Advocates have veto power on experience | 2025-12-13 |
| Full design trio (SD+UX+UI) | Terminal UI still serves humans, needs design thinking | 2025-12-13 |
| Three-tier CLAUDE.md | 85% token reduction per session | 2025-12-13 |
| Rich for UI | Terminal-native, no web server needed | Original |
| uv for packages | Fast, reliable Python package management | Original |

## Next Steps

1. [x] Complete agent definitions (all 6 agents configured)
2. [x] Create context reference cards
3. [x] Set up MCP server launcher
4. [ ] Configure pre-commit hooks
5. [ ] Write initial test suite
6. [ ] Run Service Designer to map developer journey
7. [ ] Run UX Designer to optimize terminal interaction
8. [ ] Run UI Designer to define Rich styling

## File Structure Map

```
.claude/
├── project-context.md    # This file
├── agents/
│   ├── service-designer.md  # Human Advocate: journey mapping
│   ├── ux-designer.md       # Human Advocate: interaction design
│   ├── ui-designer.md       # Human Advocate: visual design
│   ├── architect.md         # Technical: system design
│   ├── engineer.md          # Technical: implementation
│   └── tester.md            # Technical: testing
├── reference/
│   └── development-rules.md  # Full development rules
└── commands/
    └── (custom slash commands)

docs/
├── context-refs/
│   ├── README.md
│   ├── architecture-quick.md
│   └── development-quick.md
└── shared-docs/
    └── 00-best-practices/
        ├── 01-human-centered-development.md  # Foundational philosophy
        ├── 02-claude-code-configuration.md   # AI environment setup
        └── ...
```

## Session Handoff Notes

**Human-Centered Environment Established:**
- Full design trio configured (Service Designer → UX Designer → UI Designer)
- Human Advocates have veto power on experience decisions
- Technical Implementers serve the human-centered vision
- Workflow: SD+Architect collaborate first, then parallel tracks, then integration

**Next Focus Areas:**
1. Run design agents to establish the developer experience vision
2. Then implement based on design agent outputs
3. Always route experience decisions through Human Advocates

**Key Philosophy:**
> "Software exists to serve humans. Every technical decision must answer to the human experience it enables."

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.