nanogpt

Educational GPT implementation in ~300 lines. Reproduces GPT-2 (124M) on OpenWebText. Clean, hackable code for learning transformers. By Andrej Karpathy. Perfect for understanding GPT architecture from scratch. Train on Shakespeare (CPU) or OpenWebText (multi-GPU).

Quality

77%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/ml-training/nanogpt/SKILL.md

Quality

Content

72%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

Highly actionable, well-structured content with excellent progressive disclosure, held back by duplicated quick-start/workflow content and the absence of validation checkpoints in the heavier training workflows.

Suggestions

Collapse the Quick start Shakespeare block into Workflow 1 (or vice versa) to remove the verbatim duplication and save tokens.

Add validation/verification steps to the heavier workflows — e.g. after `prepare.py` confirm train.bin/val.bin sizes, and after training check val loss / sample before declaring success.

Drop or relocate time-sensitive vanity metrics (e.g. '⭐ 48,000+') from the Resources section to keep the body evergreen.

Dimension	Reasoning	Score
Conciseness	Mostly efficient code and commands, but the Quick start duplicates the Shakespeare commands repeated verbatim in Workflow 1, and sections like 'Simplicity advantages' and 'Resources' (with star counts) pad the body; it could be tightened to earn the top anchor.	2 / 3
Actionability	Provides fully executable, copy-paste-ready commands and complete config snippets for prepare/train/sample across Shakespeare, GPT-2, fine-tuning, and custom datasets — matching the executable top anchor.	3 / 3
Workflow Clarity	Workflows are clearly sequenced with numbered Step 1/2/3, but heavy/batch operations (8-GPU 4-day training, custom dataset prep) have no validation or verification checkpoints; per guidelines missing validation in such workflows caps clarity at 2.	2 / 3
Progressive Disclosure	SKILL.md is a concise overview with well-signaled one-level-deep references to real bundle files ([references/architecture.md], [references/training.md], [references/data.md]), with content appropriately split and easy to navigate.	3 / 3
	Total	10 / 12 Passed

Description

82%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific description with good trigger terms and a clear niche, but it omits an explicit 'Use when...' clause so the triggering guidance is only implied rather than stated.

Suggestions

Add an explicit 'Use when...' clause, e.g. 'Use when learning GPT/transformer internals, reproducing GPT-2, or training small character-level models from scratch.'

Trim the editorializing ('Perfect for understanding GPT architecture from scratch.') which is mildly fluffy; let the concrete capabilities carry the description.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions in third person — 'Reproduces GPT-2 (124M) on OpenWebText', 'Train on Shakespeare (CPU) or OpenWebText (multi-GPU)', '~300 lines' — matching the top anchor for specific concrete actions.	3 / 3
Completeness	It clearly answers 'what' (educational GPT implementation reproducing GPT-2) but lacks any explicit 'Use when...' trigger clause, so 'when' is only implied; per guidelines a missing explicit trigger caps completeness at 2.	2 / 3
Trigger Term Quality	Covers natural terms a user would actually say — 'GPT', 'GPT-2', 'transformers', 'Shakespeare', 'OpenWebText', 'multi-GPU' — giving good coverage of likely request phrasings.	3 / 3
Distinctiveness Conflict Risk	The nanoGPT / educational-from-scratch niche is distinct with specific triggers (GPT-2 124M, Shakespeare, OpenWebText) unlikely to conflict with other skills.	3 / 3
	Total	11 / 12 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/ml-training/nanogpt/SKILL.md
Commit: 52845c3

Reviewed: about 13 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.