A command for acquiring or auditing project datasets. It stores downloaded archives separately from tracked licence and usage information.

Use it to select a dataset location, approve public or restricted corpora, create licence sidecars, fetch data, or audit datasets already in a project.

It helps check permissions, provenance, task relevance, and evaluation-set overlap before data is downloaded or used.

dataset — Command by stellarshenson

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods

npx agentmods add commands/stellarshenson/claude-code-plugins/dataset

Clone the repo

git clone --depth 1 https://github.com/stellarshenson/claude-code-plugins

Per session 32 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 613 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5	$0.00032	$0.00613
Opus 5	$0.00016	$0.00307
Sonnet 5	$0.00006	$0.00123
Haiku 4.5	$0.00003	$0.00061

Measured 2d ago against content hash d3e000827d7f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

dataset scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/datascience/commands/dataset.md · 24 lines

How it starts

The opening of the file, as written. The whole thing — 24 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Dataset

Read the datascience:dataset skill first - it is the single source of truth for the two-artifact layout, the admission gate, the sidecar shape, and the fetcher contract. Do NOT duplicate its content here. The examples/ sidecars and fetcher skeleton win on any conflict.

What to do

Read the datascience:dataset skill, then the closest examples/ artifact - dataset-ragtruth.md for a public corpus, dataset-edgar-restricted.md for a restricted one, fetch_datasets.py for the fetcher skeleton
Location gate (before creating anything) - propose data/external/<task>-datasets/ plus the alternatives visible on disk; WAIT for the answer, then reuse the confirmed location for the rest of that task without asking again
Admission gate before any download - licence permits the intended use (commercial, training and redistribution are three separate permissions), the corpus carries what the task consumes, the task-shape mapping writes as a one-line rewrite, no provenance overlap with the evaluation set. Name which filter a rejected corpus failed
Sidecar first - the spec dict in scripts/fetch_<task>_datasets.py is the single source of truth; --dry-run renders every sidecar and fetches nothing, so licences and sizes are read before a byte costs bandwidth. Never hand-write or hand-edit a sidecar
Fetch - named args fetch one corpus, no args fetch all; a failed corpus prints SKIP <id>: <error> and continues; stage, archive, remove staging; the archive contains its own sidecar
Gitignore - whitelist the sidecars, ignore everything else under the folder; verify with git check-ignore -v on one archive and one sidecar before the first commit
Restricted corpus - the three extra sidecar bullets (restriction clauses, reason, exclusion mechanism); no archive at all where redistribution is forbidden; credentials come from the environment or the vault at fetch time - never the spec, the sidecar, or the repo

Audit mode

Read the full file on GitHub · 24 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2d ago First seen · 24 lines · 32 tokens per session scan A d3e000827d7f

Subscribe to this mod's changes

dataset is a command published in the GitHub repository stellarshenson/claude-code-plugins (3 stars, last pushed 3d ago), licensed MIT. It adds 32 tokens to every session and 613 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Data and AI Research

dataset

Getting it into your agent

What it costs to keep this loaded

Grade A, and why

Nothing flagged

How it starts

Dataset

What to do

Audit mode

What this file has done since we first saw it

Other commands, from other repositories

awesome-chatgpt

flow-nexus-neural

openrouter-setup

laravel-ai-sdk

data-scientist

audit-prompt