| name | para-init |
| description | Initialize PARA-Programming structure in the current project. Use when setting up a new repo for PARA workflow, creating context/ directory, or bootstrapping AGENTS.md. |
| model | haiku |
| effort | low |
Initialize PARA-Programming structure in the current project.
Usage
para-init
para-init --template=basic # Minimal project AGENTS.md (default)
para-init --template=full # Comprehensive project AGENTS.md
What It Does
- Set up global methodology file at
~/.agents/AGENTS.md (copied from resources/AGENTS.md if missing; never overwrites existing).
- Create context directory structure:
mkdir -p context/{data,plans,summaries,archives,servers}
- Create
context/context.md seeded from assets/context-template.md:
# Current Work Summary
Ready to start first task.
---
```json
{
"active_context": [],
"completed_summaries": [],
"last_updated": "TIMESTAMP"
}
```
See references/context-schema.md for the full field reference.
- Create project
AGENTS.md (if missing) from assets/agents-basic-template.md (default, --template=basic) or assets/agents-full-template.md (--template=full).
- Update
.gitignore to include .para-worktrees/:
- If
.gitignore exists, check for .para-worktrees/ entry; append if missing.
- If
.gitignore does not exist, create it with .para-worktrees/ as its content.
- This prevents worktree directories from being tracked by git.
Graceful Degradation (partial-install fallback)
If references/context-schema.md is not available in this install (single-skill copy without the bundled resource), the minimal required context.md JSON schema is:
{
"active_context": [],
"completed_summaries": [],
"last_updated": "ISO-8601"
}
Required fields: active_context (string[]), completed_summaries (string[]), last_updated (ISO 8601 string). Optional fields include research_docs (string[]), worktree_path, execution_branch, execution_started, phased_execution, and workflow — all documented in the full schema.
Success Output
After initialization, display:
PARA-Programming Structure Initialized
context/
├── archives/ # Historical context snapshots
├── data/ # Input files, payloads, datasets
├── plans/ # Pre-work planning documents
├── servers/ # MCP tool wrappers
├── summaries/ # Post-work reports
└── context.md # Active session context
.para-worktrees/ # Git worktree isolation (gitignored)
Files created/updated:
- ~/.agents/AGENTS.md (global methodology, if it didn't exist)
- context/context.md (fresh context file)
- AGENTS.md (project-specific context, if it didn't exist)
- .gitignore (added .para-worktrees/ entry)
Next steps:
1. Edit AGENTS.md with your project-specific context
2. Create your first plan: use the `para-plan` skill with `<task-description>`
3. Check status: use the `para-status` skill
4. Get help: use the `para-help` skill