---
name: Directory Management
slug: directory-management
category: Automation
description: Directory Management resolves a project name and creates a standard folder structure for plans, scripts, notebooks, manifests, and agent memory. Use it when starting a new project or resuming an existing one with a PLAN.md.
github: "https://github.com/awslabs/agent-plugins/tree/main/plugins/sagemaker-ai/skills/directory-management"
language: Python
stars: 867
forks: 145
install: "npx degit https://github.com/awslabs/agent-plugins/tree/main/plugins/sagemaker-ai/skills/directory-management ~/.claude/skills/directory-management"
installs_to: ~/.claude/skills/directory-management
source_path: plugins/sagemaker-ai/skills/directory-management/SKILL.md
collection_size: 25
category_size: 1523
collection_url: "https://dirskills.com/collections/awslabs/agent-plugins"
added: 2026-08-22T05:21:36.496Z
last_synced: 2026-08-22T05:21:36.496Z
canonical_url: "https://dirskills.com/skills/directory-management"
---

# Directory Management

Directory Management resolves a project name and creates a standard folder structure for plans, scripts, notebooks, manifests, and agent memory. Use it when starting a new project or resuming an existing one with a PLAN.md.

**Install:**

```bash
npx degit https://github.com/awslabs/agent-plugins/tree/main/plugins/sagemaker-ai/skills/directory-management ~/.claude/skills/directory-management
```

## README

# Directory Management

## Project Setup

Before any work begins, resolve the project name:

1. If the project name is already known from conversation context, use it.
2. Otherwise, scan for existing `*/PLAN.md` files in the current directory. If found, ask the user if they are resuming an existing project and load that `PLAN.md` into context.
3. If no existing projects are found, recommend a ≤64-char lowercase slug based on what you know from the conversation (only `[a-z0-9-]`), or ask directly if there isn't enough context. Present the recommended name and wait for user confirmation.

Once project name is resolved:

1. Create and/or use the `<experiment-name>/` directory using the confirmed name for storing all the artifacts

## Directory Structure

When working with the agent, all generated files are organized under an project directory.

```
<project-name>/
├── specs/  
│   ├── PLAN.md             # Your customization plan
├── scripts/                # Generated Python scripts
│   ├── <project-name>_transform_fn.py
├── notebooks/              # Generated Jupyter notebooks
│   ├── <project-name>.ipynb
├── manifests/              # Machine-readable outputs (JSON)
└── agent_memory/           # Session persistence (git-ignored)
    └── session-notes.md    # Progress, artifacts, next steps
```