Back to skill

Security audit

machine-learning-engineer

Security checks for vulnerabilities and agentic risk

Overview

The skill provides the advertised machine-learning and AI-development help, with ordinary precautions needed for installation and private evaluation data.

Install only the components you need, preferably project-local and from reviewed, pinned sources. Before evaluating private documents or production logs, approve the model provider, tracking destination, access controls, and retention settings; local logging is an option. Bound background evaluation costs. Verify class-distribution drift independently because the CSV checker’s calculation is flawed. Some evaluation helper scripts are not bundled, and future remote dependency versions remain outside this review.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (61)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

Most installed skills correspond to the stated ML focus or the explicitly listed agent-related triggers. Installation is a supporting implementation detail and is not itself a mismatch. However, the bundle also installs mcp-server-generation, extending its scope to MCP server development without declaring that capability. The supplied commands establish installation only; they do not demonstrate actual server generation, model training, or agent evaluation.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · AGENTS.md (reported line 9)May include surrounding context.

md
| 环节 | 技能 | 文件 | 用途 |
|------|------|------|------|
| ML 基础 | machine-learning | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | ML 全流程开发:JAX、特征工程、模型选择、评估 |
| 深度学习 | deep-learning-pytorch | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | PyTorch 深度学习:Transformers、扩散模型、LLM |
| 视觉 | computer-vision-opencv | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉:OpenCV、图像/视频处理、目标检测 |
| 工程 | pytorch-patterns | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | PyTorch 工程实践:训练管线、模型架构、数据加载 |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · AGENTS.md (reported line 9)May include surrounding context.

md
| 环节 | 技能 | 文件 | 用途 |
|------|------|------|------|
| ML 基础 | machine-learning | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | ML 全流程开发:JAX、特征工程、模型选择、评估 |
| 深度学习 | deep-learning-pytorch | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | PyTorch 深度学习:Transformers、扩散模型、LLM |
| 视觉 | computer-vision-opencv | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉:OpenCV、图像/视频处理、目标检测 |
| 工程 | pytorch-patterns | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | PyTorch 工程实践:训练管线、模型架构、数据加载 |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 41)May include surrounding context.

md
| ML 管线设计(训练循环/数据加载/检查点) | [references/ml-pipeline-design.md](references/ml-pipeline-design.md) | ml 管线 训练循环 数据加载 检查点 分布式 训练 |
| 计算机视觉:OpenCV+PyTorch 图像/视频处理、检测、分割 | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉 opencv 图像 视频 目标检测 分割 特征提取 |
| PyTorch 深度学习:Transformers、扩散模型、LLM 开发 | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | pytorch 深度学习 transformer 扩散模型 llm gradio 部署 |
| ML 全流程:JAX、特征工程、模型选择与评估 | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | 机器学习 jax scikit-learn 特征工程 模型选择 评估 调优 |
| PyTorch 工程实践:训练管线、数据加载、架构设计 | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | pytorch 工程 训练管线 数据加载 模型架构 可复现 |
| 用 Claude API / Agent SDK 构建 AI 应用与 Agent:Messages API、流式、工具调用、Agent 编排(代码骨架/决策树/踩坑见 playbook) | [skills/claude-agent-sdk/SKILL.md](skills/claude-agent-sdk/SKILL.md) [references/claude-api-playbook.md](references/claude-api-playbook.md) | claude-api anthropic agent-sdk 工具调

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 41)May include surrounding context.

md
数据加载/检查点) | [references/ml-pipeline-design.md](references/ml-pipeline-design.md) | ml 管线 训练循环 数据加载 检查点 分布式 训练 |
| 计算机视觉:OpenCV+PyTorch 图像/视频处理、检测、分割 | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉 opencv 图像 视频 目标检测 分割 特征提取 |
| PyTorch 深度学习:Transformers、扩散模型、LLM 开发 | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | pytorch 深度学习 transformer 扩散模型 llm gradio 部署 |
| ML 全流程:JAX、特征工程、模型选择与评估 | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | 机器学习 jax scikit-learn 特征工程 模型选择 评估 调优 |
| PyTorch 工程实践:训练管线、数据加载、架构设计 | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | pytorch 工程 训练管线 数据加载 模型架构 可复现 |
| 用 Claude API / Agent SDK 构建 AI 应用与 Agent:Messages API、流式、工具调用、Agent 编排(代码骨架/决策树/踩坑见 playbook) | [skills/claude-agent-sdk/SKILL.md](skills/claude-agent-sdk/SKILL.md) [references/claude-api-playbook.md](references/claude-api-playbook.md) | claude-api anthropic agent-sdk 工具调用 构建agent 流式 ll

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/experiment-reproducibility.md (reported line 47)May include surrounding context.

md
本 playbook 提供「实验可复现」的方法论(L2 决策层),落地到以下子技能:

- [machine-learning 子技能](../skills/machine-learning/SKILL.md) — 基线建模与评估的可复现维度(种子 / 数据版本)
- [deep-learning-pytorch 子技能](../skills/deep-learning-pytorch/SKILL.md) — 训练脚本的种子、环境与数据版本化
- [pytorch-patterns 子技能](../skills/pytorch-patterns/SKILL.md) — 训练管线的可复现清单与工程实践

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/ml-pipeline-design.md (reported line 64)May include surrounding context.

md
- [deep-learning-pytorch 子技能](../skills/deep-learning-pytorch/SKILL.md) — PyTorch 训练循环、分布式扩容与部署
- [pytorch-patterns 子技能](../skills/pytorch-patterns/SKILL.md) — 训练管线、数据加载与架构设计最佳实践
- [machine-learning 子技能](../skills/machine-learning/SKILL.md) — 从基线到深度学习的管线衔接

边界:本篇给管线权衡与瓶颈判断,不替 `skills/` 下的子技能选框架或编写管线代码。

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/model-evaluation.md (reported line 58)May include surrounding context.

md
本 playbook 提供「按任务选指标、防泄漏验证」的评估方法(L2 决策层),落地到以下子技能:

- [machine-learning 子技能](../skills/machine-learning/SKILL.md) — 模型选择与评估的核心方法
- [deep-learning-pytorch 子技能](../skills/deep-learning-pytorch/SKILL.md) — 深度学习模型的验证与调优
- [computer-vision-opencv 子技能](../skills/computer-vision-opencv/SKILL.md) — 视觉任务的指标与验证策略
- [pytorch-patterns 子技能](../skills/pytorch-patterns/SKILL.md) — 可复现的评估实验设计

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · skills/machine-learning/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/machine-learning-engineer
  path: skills/machine-learning/SKILL.md
  url: https://skills.sh/mindrally/skills/machine-learning
  version: 1.0.0
  updated: 2026-07-30

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · AGENTS.md (reported line 10)May include surrounding context.

md
| 环节 | 技能 | 文件 | 用途 |
|------|------|------|------|
| ML 基础 | machine-learning | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | ML 全流程开发:JAX、特征工程、模型选择、评估 |
| 深度学习 | deep-learning-pytorch | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | PyTorch 深度学习:Transformers、扩散模型、LLM |
| 视觉 | computer-vision-opencv | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉:OpenCV、图像/视频处理、目标检测 |
| 工程 | pytorch-patterns | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | PyTorch 工程实践:训练管线、模型架构、数据加载 |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · AGENTS.md (reported line 10)May include surrounding context.

md
| 环节 | 技能 | 文件 | 用途 |
|------|------|------|------|
| ML 基础 | machine-learning | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | ML 全流程开发:JAX、特征工程、模型选择、评估 |
| 深度学习 | deep-learning-pytorch | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | PyTorch 深度学习:Transformers、扩散模型、LLM |
| 视觉 | computer-vision-opencv | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉:OpenCV、图像/视频处理、目标检测 |
| 工程 | pytorch-patterns | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | PyTorch 工程实践:训练管线、模型架构、数据加载 |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 40)May include surrounding context.

md
| 模型评估框架(指标/验证/基线/泄漏) | [references/model-evaluation.md](references/model-evaluation.md) | 模型评估 指标 验证 基线 数据泄漏 校准 |
| ML 管线设计(训练循环/数据加载/检查点) | [references/ml-pipeline-design.md](references/ml-pipeline-design.md) | ml 管线 训练循环 数据加载 检查点 分布式 训练 |
| 计算机视觉:OpenCV+PyTorch 图像/视频处理、检测、分割 | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉 opencv 图像 视频 目标检测 分割 特征提取 |
| PyTorch 深度学习:Transformers、扩散模型、LLM 开发 | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | pytorch 深度学习 transformer 扩散模型 llm gradio 部署 |
| ML 全流程:JAX、特征工程、模型选择与评估 | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | 机器学习 jax scikit-learn 特征工程 模型选择 评估 调优 |
| PyTorch 工程实践:训练管线、数据加载、架构设计 | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | pytorch 工程 训练管线 数据加载 模型架构 可复现 |
| 用 Claude API / Agent SDK 构建 AI 应用与 Agent:Messages API、流式、工具调用、Agent 编排(代码骨架/决策树/踩坑见 playbook) | [skills/claude-agent-sdk/SKILL.md](skills/claude-agent-sdk/SK

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 40)May include surrounding context.

md
| 模型评估框架(指标/验证/基线/泄漏) | [references/model-evaluation.md](references/model-evaluation.md) | 模型评估 指标 验证 基线 数据泄漏 校准 |
| ML 管线设计(训练循环/数据加载/检查点) | [references/ml-pipeline-design.md](references/ml-pipeline-design.md) | ml 管线 训练循环 数据加载 检查点 分布式 训练 |
| 计算机视觉:OpenCV+PyTorch 图像/视频处理、检测、分割 | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉 opencv 图像 视频 目标检测 分割 特征提取 |
| PyTorch 深度学习:Transformers、扩散模型、LLM 开发 | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | pytorch 深度学习 transformer 扩散模型 llm gradio 部署 |
| ML 全流程:JAX、特征工程、模型选择与评估 | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | 机器学习 jax scikit-learn 特征工程 模型选择 评估 调优 |
| PyTorch 工程实践:训练管线、数据加载、架构设计 | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | pytorch 工程 训练管线 数据加载 模型架构 可复现 |
| 用 Claude API / Agent SDK 构建 AI 应用与 Agent:Messages API、流式、工具调用、Agent 编排(代码骨架/决策树/踩坑见 playbook) | [skills/claude-agent-sdk/SKILL.md](skills/claude-agent-sdk/SK

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/experiment-reproducibility.md (reported line 48)May include surrounding context.

md
本 playbook 提供「实验可复现」的方法论(L2 决策层),落地到以下子技能:

- [machine-learning 子技能](../skills/machine-learning/SKILL.md) — 基线建模与评估的可复现维度(种子 / 数据版本)
- [deep-learning-pytorch 子技能](../skills/deep-learning-pytorch/SKILL.md) — 训练脚本的种子、环境与数据版本化
- [pytorch-patterns 子技能](../skills/pytorch-patterns/SKILL.md) — 训练管线的可复现清单与工程实践

边界:本篇只给可复现维度的检查建议,不替 `skills/` 下的子技能决定实验设计或编写代码。

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/ml-pipeline-design.md (reported line 62)May include surrounding context.

md
本 playbook 提供「训练管线与工程化」的设计模式(L2 决策层),落地到以下子技能:

- [deep-learning-pytorch 子技能](../skills/deep-learning-pytorch/SKILL.md) — PyTorch 训练循环、分布式扩容与部署
- [pytorch-patterns 子技能](../skills/pytorch-patterns/SKILL.md) — 训练管线、数据加载与架构设计最佳实践
- [machine-learning 子技能](../skills/machine-learning/SKILL.md) — 从基线到深度学习的管线衔接

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/model-evaluation.md (reported line 59)May include surrounding context.

md
本 playbook 提供「按任务选指标、防泄漏验证」的评估方法(L2 决策层),落地到以下子技能:

- [machine-learning 子技能](../skills/machine-learning/SKILL.md) — 模型选择与评估的核心方法
- [deep-learning-pytorch 子技能](../skills/deep-learning-pytorch/SKILL.md) — 深度学习模型的验证与调优
- [computer-vision-opencv 子技能](../skills/computer-vision-opencv/SKILL.md) — 视觉任务的指标与验证策略
- [pytorch-patterns 子技能](../skills/pytorch-patterns/SKILL.md) — 可复现的评估实验设计

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · skills/deep-learning-pytorch/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/machine-learning-engineer
  path: skills/deep-learning-pytorch/SKILL.md
  url: https://skills.sh/mindrally/skills/deep-learning-pytorch
  version: 1.0.0
  updated: 2026-07-30

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · AGENTS.md (reported line 11)May include surrounding context.

md
|------|------|------|------|
| ML 基础 | machine-learning | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | ML 全流程开发:JAX、特征工程、模型选择、评估 |
| 深度学习 | deep-learning-pytorch | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | PyTorch 深度学习:Transformers、扩散模型、LLM |
| 视觉 | computer-vision-opencv | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉:OpenCV、图像/视频处理、目标检测 |
| 工程 | pytorch-patterns | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | PyTorch 工程实践:训练管线、模型架构、数据加载 |

## 适用场景

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · AGENTS.md (reported line 11)May include surrounding context.

md
|------|------|------|------|
| ML 基础 | machine-learning | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | ML 全流程开发:JAX、特征工程、模型选择、评估 |
| 深度学习 | deep-learning-pytorch | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | PyTorch 深度学习:Transformers、扩散模型、LLM |
| 视觉 | computer-vision-opencv | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉:OpenCV、图像/视频处理、目标检测 |
| 工程 | pytorch-patterns | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | PyTorch 工程实践:训练管线、模型架构、数据加载 |

## 适用场景

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 39)May include surrounding context.

md
| 实验可复现性(种子/版本/配置/数据版本) | [references/experiment-reproducibility.md](references/experiment-reproducibility.md) | 可复现 随机种子 环境版本 配置管理 数据版本 实验追踪 |
| 模型评估框架(指标/验证/基线/泄漏) | [references/model-evaluation.md](references/model-evaluation.md) | 模型评估 指标 验证 基线 数据泄漏 校准 |
| ML 管线设计(训练循环/数据加载/检查点) | [references/ml-pipeline-design.md](references/ml-pipeline-design.md) | ml 管线 训练循环 数据加载 检查点 分布式 训练 |
| 计算机视觉:OpenCV+PyTorch 图像/视频处理、检测、分割 | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉 opencv 图像 视频 目标检测 分割 特征提取 |
| PyTorch 深度学习:Transformers、扩散模型、LLM 开发 | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | pytorch 深度学习 transformer 扩散模型 llm gradio 部署 |
| ML 全流程:JAX、特征工程、模型选择与评估 | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | 机器学习 jax scikit-learn 特征工程 模型选择 评估 调优 |
| PyTorch 工程实践:训练管线、数据加载、架构设计 | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | pytorch 工程 训练管线 数据加载 模型架构 可复现 |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 39)May include surrounding context.

md
| 实验可复现性(种子/版本/配置/数据版本) | [references/experiment-reproducibility.md](references/experiment-reproducibility.md) | 可复现 随机种子 环境版本 配置管理 数据版本 实验追踪 |
| 模型评估框架(指标/验证/基线/泄漏) | [references/model-evaluation.md](references/model-evaluation.md) | 模型评估 指标 验证 基线 数据泄漏 校准 |
| ML 管线设计(训练循环/数据加载/检查点) | [references/ml-pipeline-design.md](references/ml-pipeline-design.md) | ml 管线 训练循环 数据加载 检查点 分布式 训练 |
| 计算机视觉:OpenCV+PyTorch 图像/视频处理、检测、分割 | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉 opencv 图像 视频 目标检测 分割 特征提取 |
| PyTorch 深度学习:Transformers、扩散模型、LLM 开发 | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | pytorch 深度学习 transformer 扩散模型 llm gradio 部署 |
| ML 全流程:JAX、特征工程、模型选择与评估 | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | 机器学习 jax scikit-learn 特征工程 模型选择 评估 调优 |
| PyTorch 工程实践:训练管线、数据加载、架构设计 | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | pytorch 工程 训练管线 数据加载 模型架构 可复现 |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/model-evaluation.md (reported line 60)May include surrounding context.

md
- [machine-learning 子技能](../skills/machine-learning/SKILL.md) — 模型选择与评估的核心方法
- [deep-learning-pytorch 子技能](../skills/deep-learning-pytorch/SKILL.md) — 深度学习模型的验证与调优
- [computer-vision-opencv 子技能](../skills/computer-vision-opencv/SKILL.md) — 视觉任务的指标与验证策略
- [pytorch-patterns 子技能](../skills/pytorch-patterns/SKILL.md) — 可复现的评估实验设计

边界:本篇给指标 / 验证 / 泄漏的相对判断,不替 `skills/` 下的子技能拍板上线。

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · skills/computer-vision-opencv/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/machine-learning-engineer
  path: skills/computer-vision-opencv/SKILL.md
  url: https://skills.sh/mindrally/skills/computer-vision-opencv
  version: 1.0.0
  updated: 2026-07-30

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · AGENTS.md (reported line 12)May include surrounding context.

md
| ML 基础 | machine-learning | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | ML 全流程开发:JAX、特征工程、模型选择、评估 |
| 深度学习 | deep-learning-pytorch | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | PyTorch 深度学习:Transformers、扩散模型、LLM |
| 视觉 | computer-vision-opencv | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉:OpenCV、图像/视频处理、目标检测 |
| 工程 | pytorch-patterns | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | PyTorch 工程实践:训练管线、模型架构、数据加载 |

## 适用场景

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · AGENTS.md (reported line 12)May include surrounding context.

md
| ML 基础 | machine-learning | [skills/machine-learning/SKILL.md](skills/machine-learning/SKILL.md) | ML 全流程开发:JAX、特征工程、模型选择、评估 |
| 深度学习 | deep-learning-pytorch | [skills/deep-learning-pytorch/SKILL.md](skills/deep-learning-pytorch/SKILL.md) | PyTorch 深度学习:Transformers、扩散模型、LLM |
| 视觉 | computer-vision-opencv | [skills/computer-vision-opencv/SKILL.md](skills/computer-vision-opencv/SKILL.md) | 计算机视觉:OpenCV、图像/视频处理、目标检测 |
| 工程 | pytorch-patterns | [skills/pytorch-patterns/SKILL.md](skills/pytorch-patterns/SKILL.md) | PyTorch 工程实践:训练管线、模型架构、数据加载 |

## 适用场景