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

74%

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

Fix and improve this skill with Tessl

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

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

Quality

Content

61%Weight 40%Scale 1-5

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

The body is a well-structured, actionable overview that pushes detail into real reference files with consistent navigation. Its main weaknesses are duplicated reference listings between inline sections and the Resources section, and the absence of explicit validation checkpoints in the long-running training workflows.

Suggestions

Consolidate the per-capability reference bullet lists into a single Resources section to remove the near-verbatim duplication (e.g. lines 66-72 vs 301-310).

Add explicit validation/verification checkpoints to the training and custom-environment workflows (e.g. "Validate the environment with the test utility before scaling vectorization").

Replace undefined placeholders in the policy and training examples (`my_policy`, `num_iterations`, `obs_dim`, `num_actions`) with minimal concrete definitions so snippets run as-is.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient and assumes Claude's competence, but it duplicates content: each capability section ends with a "For X, read references/X.md for:" bullet list, and the Resources section (lines ~299-355) repeats those same bullet lists almost verbatim, plus generic filler like "Achieve maximum throughput with optimized parallel simulation" — fitting the "mostly efficient but includes some unnecessary explanation or could be tightened" anchor; not a 4 because the redundancy is more than minor.	3 / 5
Actionability	Provides concrete, copy-paste-ready code across training, environments, policies, integration, and common use cases (e.g. CLI `puffer train ...`, `PuffeRL(...)` loop, `pufferlib.make(...)`), matching the "mostly executable guidance; concrete code or commands with minor gaps" anchor; not a 5 because some examples use undefined placeholders (`my_policy`, `num_iterations`, `obs_dim`, `num_actions`).	4 / 5
Workflow Clarity	Several clearly sequenced numbered workflows exist ("For Training Existing Environments", "For Creating Custom Environments", etc.) with next-step references, but long-running batch training operations lack explicit validation/verification checkpoints, which caps workflow clarity at 3 per the rubric's batch-operations guideline; not a 4 because the validation gap is material rather than minor.	3 / 5
Progressive Disclosure	Clear overview body with well-signaled, one-level-deep references to real files (`references/.md`, `scripts/.py`, all verified to exist) and consistent navigation cues, matching the "good structure; most content appropriately placed; minor organization gaps" anchor; not a 5 because the inline reference bullet lists are duplicated in the Resources section, slightly muddying navigation.	4 / 5
	Total	14 / 20 Passed

Description

87%Weight 40%Scale 1-5

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

The description is strong: it states a concrete capability set, gives explicit "Use when" triggers with named frameworks, and even routes users away to a competitor for the wrong fit. Its only weak spot is a marketing-style speedup claim and a few missing natural synonyms.

Dimension	Reasoning	Score
Specificity	Names multiple concrete capabilities ("fast parallel training, vectorized environments, multi-agent systems, or integration with game environments (Atari, Procgen, NetHack)") and an explicit alternative-routing clause, fitting the "lists several specific actions; minor gaps" anchor; not a 5 because the "2-10x speedups" claim is marketing framing rather than an additional concrete capability action.	4 / 5
Completeness	Explicitly answers both what ("High-performance reinforcement learning framework optimized for speed and scale") and when ("Use when you need fast parallel training...") with concrete trigger phrases plus a negative-routing clause, matching the top anchor; not below because both elements are clearly and explicitly present.	5 / 5
Trigger Term Quality	Strong natural trigger coverage ("fast parallel training", "vectorized environments", "multi-agent systems", "Atari, Procgen, NetHack", "stable-baselines3") matching the "good keyword coverage; a few natural terms missing" anchor; not a 5 because synonyms like "RL", "PPO", or "reinforcement learning" phrasings users might say are absent.	4 / 5
Distinctiveness Conflict Risk	Clear niche (high-performance RL), specific framework names, and an explicit steer-away to stable-baselines3 for the adjacent use case, giving a distinct trigger profile with minimal conflict risk per the top anchor.	5 / 5
	Total	18 / 20 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: 5f2a37b

Reviewed: 5 days 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.