pufferlib

High-performance reinforcement learning framework optimized for speed and scale. Use when you need fast parallel training, vectorized environments, multi-agent systems, or integration with game environments (Atari, Procgen, NetHack). Achieves 2-10x speedups over standard implementations. For quick prototyping or standard algorithm implementations with extensive documentation, use stable-baselines3 instead.

Quality

86%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

The canonical home for this skill is pufferlib in K-Dense-AI/scientific-agent-skills

Quality

Content

72%

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

A well-structured, highly actionable skill body with excellent progressive disclosure (all referenced files exist and stay one level deep) and copy-paste-ready code. Its main weaknesses are conciseness (the Resources section duplicates inline bullet lists) and implicit rather than explicit validation checkpoints in the workflows.

Suggestions

Remove or condense the standalone "Resources" section (lines 281–354): its per-file bullet lists duplicate the "read references/X.md for:" bullets already shown inline in each capability section, costing ~70 lines.

Add explicit validation checkpoints to the Quick Start workflows (e.g. a numbered "Verify the environment passes its test suite before vectorizing/training" step) so verification is a gate rather than a tip.

Trim "Tips for Success" to the few that add non-reference guidance (e.g. profiling early, progressive optimization); drop items that merely restate reading the references.

Dimension	Reasoning	Score
Conciseness	The body is code-forward and largely efficient, but the ~70-line "Resources" section (lines 281–354) restates the same per-reference bullet lists already given inline at each capability section, and the 10-item "Tips for Success" partly reiterates reference guidance — pushing it from lean to "could be tightened."	2 / 3
Actionability	Provides copy-paste-ready, executable guidance throughout: CLI commands (`puffer train procgen-coinrun --train.device cuda`), `torchrun` distributed invocation, full training loops, PufferEnv and Policy class skeletons with `layer_init`, integration snippets, and an install command — matching the fully-executable anchor.	3 / 3
Workflow Clarity	Numbered Quick Start workflows give a clear sequence for training, custom-env, policy, and performance paths, but validation/verification checkpoints are mostly implicit (e.g. "test environment locally" appears as a step but not as an explicit verify-before-proceed gate), fitting the sequence-present-but-checkpoints-implicit anchor.	2 / 3
Progressive Disclosure	SKILL.md is a genuine overview; each capability ends with a clearly signaled one-level-deep pointer (e.g. "read `references/training.md`") to real, verified bundle files, and the reference files do not nest-reference one another — matching the clear-overview-with-well-signaled-one-level-deep-references anchor.	3 / 3
	Total	10 / 12 Passed

Description

100%

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, third-person description that states concrete capabilities, provides explicit "Use when" triggers with natural RL vocabulary, and adds a clear disambiguating anti-trigger. It cleanly satisfies all four dimensions at the top of the scale.

Dimension	Reasoning	Score
Specificity	Names multiple concrete capabilities — "fast parallel training, vectorized environments, multi-agent systems, or integration with game environments (Atari, Procgen, NetHack)" — matching the multiple-specific-actions anchor, and is not vague or abstract.	3 / 3
Completeness	Explicitly answers both what (high-performance RL training/vectorization/multi-agent/integration) and when via a clear "Use when you need..." clause, and adds an explicit anti-trigger pointing to stable-baselines3 — meeting the what-AND-when-with-explicit-triggers anchor.	3 / 3
Trigger Term Quality	Covers natural terms users would actually say ("parallel training", "vectorized environments", "multi-agent systems", "Atari", "Procgen", "NetHack") plus the domain phrase "reinforcement learning framework", giving good keyword coverage rather than jargon-only.	3 / 3
Distinctiveness Conflict Risk	Occupies a clear niche (high-performance, large-scale RL with specific game environments) and explicitly disambiguates from stable-baselines3, making wrong-skill triggering unlikely.	3 / 3
	Total	12 / 12 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

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

Reviewed: about 12 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.