OpenParable
Skip to content
← Back to skills

Agentsociety Create Dataset

Apps & Automation
agentsociety-create-dataset

Agentsociety Create Dataset is an Agent Skill for apps & automation published in the tsinghua-fib-lab/AgentSociety repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Use when packaging, validating, or publishing a dataset to the AgentSociety platform for later sharing or reuse.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/tsinghua-fib-lab/AgentSociety/blob/HEAD/extension/skills/agentsociety-create-dataset/v1.0.0/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/agentsociety-create-dataset, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Create Dataset

Overview

Create, validate, package, and upload datasets to the AgentSociety platform. Produces a ZIP package conforming to the platform's schema for sharing or admin review.

Use the Python interpreter from .env. See CLAUDE.md for setup.

When to Use

  • User wants to publish or share a dataset on the AgentSociety platform
  • Packaging survey results, agent profiles, or simulation outputs for upload
  • User says "create dataset", "upload dataset", or "publish data"

Do NOT use when:

  • User wants to download or browse datasets (use agentsociety-use-dataset)
  • User is working with experiment config files only (use experiment-config)

Quick Reference

CommandAuthDescription
loginNoCasdoor Device Code Flow authentication
logoutNoClear saved credentials
init <name>NoCreate dataset directory structure
validate <path>NoValidate dataset dir or ZIP
pack <dir>NoPackage directory into ZIP
upload <zip>YesUpload to AgentSociety platform
submit <dataset_id>YesSubmit for admin review

Run commands from the workspace root through .agentsociety/bin/ags.py.

Workflow

digraph create_dataset_flow {
    rankdir=LR;
    node [shape=box, style=filled, fillcolor="#E8F4FD"];
    login [label="login\nif upload needed"];
    init [label="init dataset dir"];
    add [label="add data files"];
    meta [label="edit README.md\nand dataset.json"];
    validate [label="validate"];
    pack [label="pack zip"];
    upload [label="upload"];
    submit [label="submit for review"];

    login -> init -> add -> meta -> validate -> pack -> upload -> submit;
}

Dataset Format

A valid dataset is a directory containing:

<name>/
├── README.md         # Required. Describes the dataset.
├── dataset.json      # Required. Metadata (id, name, description, category, etc.)
└── data/             # Recommended. Dataset files (CSV, JSON, etc.)

Constraints:

  • id must match ^[a-z0-9_-]+$ and be unique on the platform
  • category: one of agent_profiles, surveys, experiments, literature, simulation_results, other
  • Total package size: max 5TB (files ≥200MB use multipart direct-to-OSS upload)

For the full dataset.json schema, see references/dataset-schema.md.

Common Mistakes

MistakeFix
Using invalid slug format in idMust match ^[a-z0-9_-]+$ (lowercase, digits, hyphens, underscores)
Exceeding 5TB package sizeSplit into multiple datasets or compress data files
Missing README.md or dataset.jsonBoth are required at the package root
Not running validate before packAlways validate then pack to catch errors early
Skipping login before upload/submitRun login first; credentials persist in ~/.agentsociety/

Pipeline Position

Predecessors: None (standalone) Successors: experiment-config (when external data needed), agentsociety-use-dataset (for download after upload)