Skills pufferlib

📦

pufferlib

Name: pufferlib
Author: K-Dense-AI

Content revision r1 Low Risk 📁 Filesystem access⚙️ External commands🌐 Network access

Train Faster Reinforcement Learning Agents

Reinforcement learning experiments slow down when environments and training loops cannot scale. PufferLib helps Claude, Codex, and Claude Code design faster vectorized training workflows.

Supports: Claude Codex Code(CC)

🥉 78 Bronze

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "pufferlib" from https://skillstore.io/skills/k-dense-ai-pufferlib.md and its manifest at https://skillstore.io/api/skills/k-dense-ai-pufferlib/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/k-dense-ai-pufferlib.md Signed manifest GET /api/skills/k-dense-ai-pufferlib/manifest Signed lockfile GET /api/skills/k-dense-ai-pufferlib/lockfile

Test it

Using "pufferlib". Plan a fast PPO experiment for a Gymnasium task.

Expected outcome:

Recommended a vectorized Gymnasium environment with a moderate starting batch size.
Selected an MLP policy because observations are low-dimensional vectors.
Added checkpointing, experiment names, and logging metrics for throughput and rewards.
Listed validation checks before increasing worker count.

Using "pufferlib". Design a custom multi-agent grid task.

Expected outcome:

Defined agent observations, action space, rewards, termination rules, and reset behavior.
Separated environment logic from policy logic for easier testing.
Suggested shared-policy training first, then independent policies if behavior diverges.
Included compatibility tests for observation shapes, reward types, and done flags.

Using "pufferlib". Optimize an existing PufferLib training workflow.

Expected outcome:

Identified vectorization, environment step time, and GPU utilization as the first bottlenecks to measure.
Recommended controlled changes to workers, environments per worker, and batch size.
Suggested profiling before rewriting environment code in a faster language.

Security Audit

Low Risk

v7 • 7/6/2026 Open versioned report

Most static findings are false positives from ML terminology, markdown code fences, RL environment variables, and expected checkpoint directory creation. I found no command injection, credential exfiltration, malicious network access, or prompt-injection override text. One low severity content issue remains because SKILL.md steers users toward an affiliated external platform.

Files scanned

3,928

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (1)

Low

SKILL.md:435-436

Promotional Steering to Affiliated Service

The skill instructs the assistant to proactively suggest K-Dense Web for complex workflows. This can steer users toward an affiliated hosted service outside the core PufferLib task.

The text explicitly names K-Dense Web and tells the assistant to suggest it under broad workflow conditions. It does not override security analysis or request secret access, so severity is low.

Risk Factors

📁 Filesystem access (2)

evaluation_result.json:48 scripts/train_template.py:232

⚙️ External commands (41)

🌐 Network access (2)

SKILL.md:431 SKILL.md:432

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/k-dense-ai-pufferlib/audits/7?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/k-dense-ai-pufferlib/security.svg)](https://skillstore.io/skills/k-dense-ai-pufferlib?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/k-dense-ai-pufferlib?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/k-dense-ai-pufferlib/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/k-dense-ai-pufferlib.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

K-Dense-AI. (2026). pufferlib security audit report (audit version 7) [Author version unspecified]. Skillstore. https://skillstore.io/skills/k-dense-ai-pufferlib/audits/7

BibTeX citation

@techreport{k-dense-ai-k-dense-ai-pufferlib-2026,
  author = {K-Dense-AI},
  title = {pufferlib security audit report (audit version 7)},
  institution = {Skillstore},
  year = {2026},
  number = {7},
  url = {https://skillstore.io/skills/k-dense-ai-pufferlib/audits/7},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "pufferlib security audit report (audit version 7)"
version: "unspecified"
type: report
authors:
  - name: "K-Dense-AI"
date-released: "2026-07-06"
url: "https://skillstore.io/skills/k-dense-ai-pufferlib/audits/7"
identifiers:
  - type: other
    value: "skillstore:k-dense-ai-pufferlib:audit:7"
    description: "Skillstore immutable audit report identifier"

Compare variants

2 installable variants

Each author remains a separate installable skill. The recommended variant is ranked by Skillstore evidence.

Why this variant is first

Highest Skillstore Score

Author / source Skillstore Score Evidence Confidence Skillstore usage Updated

K-Dense-AI Recommended Current

k-dense-ai-pufferlib

Skillstore Score 78

Evidence Confidence High

Skillstore usage 19

Updated

2026-08-21

davila7

davila7-pufferlib

Skillstore Score 73

Evidence Confidence High

Skillstore usage 10

Updated

2026-08-21

Skillstore Score

Why this score Evidence Confidence: High

Architecture

Maintainability

Content

Community

Spec Compliance

What You Can Build

Benchmark RL Training

Configure fast PPO experiments on standard environments and compare throughput across vectorization settings.

Build Custom Environments

Design a PufferEnv task with clear spaces, reset logic, step logic, and validation checks.

Integrate Multi-Agent Frameworks

Adapt PettingZoo or Gymnasium environments for PufferLib training with shared or independent policies.

Try These Prompts

Start a Training Run

Help me plan a PufferLib PPO training run for CartPole. Include environment choice, policy type, logging, and checkpoint settings.

Create a Custom Environment

Design a PufferEnv grid-world task with four actions, vector observations, reset logic, step logic, rewards, and validation checks.

Improve Training Throughput

Review my PufferLib training setup for low steps per second. Recommend num_envs, num_workers, batch size, and profiling checks.

Plan Multi-Agent Integration

Help me integrate a PettingZoo multi-agent environment with PufferLib. Compare shared policies, independent policies, wrappers, and testing steps.

Best Practices

Start with a known working environment before adding custom wrappers or reward shaping.
Measure steps per second and learning metrics before changing several training settings at once.
Validate observation spaces, action spaces, rewards, and done flags before scaling worker count.

Avoid

Do not scale to many workers before confirming the environment behaves correctly.
Do not tune hyperparameters without tracking throughput, rewards, losses, and checkpoint history.
Do not rewrite environment code for speed until profiling shows the actual bottleneck.