ai-projects
v1.9.0
Published
Useful CLI for AI agents to create and manage projects, tasks, skills and agents
Maintainers
Readme
AIP - AI Project Management CLI 🤖
Built for AI agents. Polished by AI agents.
A command-line tool for managing AI agent projects, tasks, and agents with structured workflows, activity logging, and automation hooks.
📚 Help System
AIP has a comprehensive built-in help system for every level:
# 🚀 New here? Start with the quick guide
aip help quickstart
# 📋 Read list of verbs by all nouns
aip --help # Same as `aip help usage`
# 🎯 List of verbs by noun
aip task --help # Same as `aip help usage task`
# 📋 Get all parameters and --options for a command
aip task create --help
# 📖 Complete API reference (comprehensive!)
aip help api
# 🛠️ Project management skill documentation (can be piped to a SKILL.md)
aip help skill
# 🔗 Hook system documentation
aip help hooks✨ Features
- Structured Projects: YAML frontmatter + markdown body
- Activity Logging: TSV-based status tracking (date, time, type, action, text)
- Automation Hooks: Pre/post hooks for create, start, update, complete
- Multi-line Support: Heredoc syntax for complex content
- Agent-Ready: Designed for autonomous AI workflows
🚀 Quick Start
# Create project with body (inline - cleanest)
aip project create "my-project" --description "..." --body $'# Goals\n- Build X\n- Solve Y'
# Create project with body (here-string for multi-line)
aip project create "my-project" --description "..." --body "$(cat <<< '# Goals
- Build X
- Solve Y')"
# Create a task in the project
aip task create my-project "first-task" --description "Get started"
# Navigate and work
cd $(aip task path first-task)
# Start working (and read in all context to stdout)
aip task start
# Log your progress
aip log append "API integration complete"
# View activity history
aip log read🏠 AIP_HOME - Project Root Directory
AIP stores all projects, tasks, and agents under $AIP_HOME.
Set it in your ~/.bashrc (or ~/.zshrc):
export AIP_HOME=/path/to/your/aip-rootAuto-detection (if AIP_HOME is not set):
AIP automatically detects the root by traversing up from your current directory looking for standard folders (projects/, skills/, agents/). This means you can cd anywhere in your project structure and AIP will find the root.
# No setup needed - just run from anywhere in your project
cd /path/to/aip-root/projects/my-project/tasks/some-task
aip task list # Works! AIP detected the root📁 Project Structure
Once AIP_HOME is determined, the structure is:
$AIP_HOME/projects/{project}/
├── main.md # Goals + body (YAML frontmatter + markdown)
├── log.tsv # Activity log (tab-separated: date, time, type, slug, action, text)
├── hooks/ # Automation: pre|post-{create,start,update,complete}
├── inputs/ # Intermediate/raw data (e.g. API responses, downloads)
├── outputs/ # Deliverables go here (e.g. reports, leads, etc.)
├── scripts/ # Automation scripts
└── tasks/{task}/
├── main.md # Task definition + body
├── log.tsv # Task activity log
└── ... # Same substructure🔧 Key Commands
Projects
aip project create "name" --description "..." --body "..."
aip project update --status done --body "$(cat final-report.md)"
aip project list
aip project path my-project # Get absolute pathTasks
aip task create my-project "task-name" --description "..."
aip task update --status in-progress
aip task list --project my-project
aip task list --search "-name"
aip task path task-name # Get absolute path
aip task start # Start working (and read in all context to stdout)Logging
aip log append "Completed the implementation"
aip log append --task my-task "Fixed the bug"
aip log read # View status historyUtilities
aip help quickstart # Get started guide
aip help skill --mode claude > ~/.claude/skills/aip/SKILL.md
aip help skill --mode hermes > ~/.hermes/skills/some_toolset/aip/SKILL.md💡 Pro Tips
# Chain commands
cd $(aip task path my-task) && aip task start
# Use here-string for complex body content
aip project update --body "$(cat <<< '# Updated Goals
- New goal 1
- New goal 2')"
# Quick status check
aip log read | tail -5🛠️ Development
This tool is meant for AI agents and will be automatically polished by AI agents based on their usage.
Suggestions and PRs welcome! 🎉
📄 License
MIT
