Agent-Skills.md

Agent Skills: OSF Datasets

Access Open Science Framework (OSF) datasets via CLI using osfclient and the OSF v2 API. Load when downloading research datasets, searching public OSF projects, or uploading data to OSF repositories.

UncategorizedID: plurigrid/asi/osf-datasets

Repository

plurigrid/asi
plurigridLicense: MIT
165

Install this agent skill to your local

pnpm dlx add-skill https://github.com/plurigrid/asi/tree/HEAD/skills/osf-datasets

Skill Files

Browse the full folder contents for osf-datasets.

Download Skill

Loading file tree…

skills/osf-datasets/SKILL.md

Skill Metadata

Name
osf-datasets
Description
Access Open Science Framework (OSF) datasets via CLI using osfclient and the OSF v2 API. Load when downloading research datasets, searching public OSF projects, or uploading data to OSF repositories.

OSF Datasets

Setup

pip install osfclient

Authentication

# Option A: Personal Access Token (preferred) -- generate at osf.io/settings/tokens
export OSF_TOKEN=<pat>

# Option B: credentials
export OSF_USERNAME=user@example.com
export OSF_PASSWORD=pass

Or per-directory .osfcli.config:

[osf]
username = user@example.com
project = abc12

Project ID

The 5-character alphanumeric slug from any osf.io/<id> URL. Example: osf.io/abc12 -> project ID is abc12.

Core Commands

| Command | Purpose | |---|---| | osf -p <id> ls | List all files | | osf -p <id> clone [dir] | Download entire project | | osf -p <id> fetch osfstorage/path.csv local.csv | Download single file | | osf -p <id> upload local.csv osfstorage/path.csv | Upload file | | osf -p <id> geturl osfstorage/path.csv | Get web URL |

Remote paths are prefixed with the storage provider, typically osfstorage/.

Direct API Access

Base URL: https://api.osf.io/v2/

Use --globoff with curl -- OSF query params use [] which curl interprets as glob ranges.

# List project files
curl -sL --globoff "https://api.osf.io/v2/nodes/<id>/files/osfstorage/" | jq '.data[].attributes.name'

# Get project metadata
curl -sL --globoff "https://api.osf.io/v2/nodes/<id>/" | jq '.data.attributes'

# Search public nodes by title
curl -sL --globoff "https://api.osf.io/v2/nodes/?filter[title]=keyword&page[size]=20" | jq '.data[] | {id, title: .attributes.title}'

# Browse subfolder (use folder ID from parent listing)
curl -sL --globoff "https://api.osf.io/v2/nodes/<id>/files/osfstorage/<folder_id>/" | jq '.data[].attributes.name'

# With auth
curl -sL --globoff -H "Authorization: Bearer $OSF_TOKEN" "https://api.osf.io/v2/nodes/<id>/files/osfstorage/"

File download links are in data[].links.download -- follow redirects with curl -L -o.

Helper Scripts

  • scripts/osf-search.sh <keyword> [page_size] -- search public OSF projects by keyword
  • scripts/osf-browse.sh <project_id> [path] -- browse project files via API (no osfclient needed)

Guidelines

  • Public projects need no auth; private ones require OSF_TOKEN or credentials
  • clone mirrors the remote directory structure locally
  • fetch requires both remote and local path arguments
  • The API paginates at 10 items by default; use ?page[size]=100 for larger listings
  • Rate limits apply -- add brief sleeps for batch operations