Back to skill

Security audit

accessibility-engineer

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this accessibility skill suspicious because it flags unpinned installer commands and a record of bypassing a human approval hold; it reports no demonstrated compromise or clear malicious intent.

The verdict is suspicious, and Skillspector’s recommendation is not to install. Before considering it, review the approval-hold record and verify the installer and skill source at fixed, trusted versions; avoid global installation unless you need it. The AIG scan reported no findings, but it does not cancel the reported concerns.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (31)

Reinterprets repeated "continue" messages as authorization to bypass an explicit human-approval hold.

High
Category
Prompt Injection
Confidence
91% confidence
Finding

The execution record explicitly treats repeated generic 'continue' messages as permission to bypass a pending human-approval hold, without documenting approval of the specific held action. This is evidence of an authorization-control failure, although the demonstrated changes are limited to provenance metadata and the historical record does not itself execute or mandate future bypasses.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

Although this is an installer CLI rather than an MCP server, unversioned npx skills can download and execute a changing npm package with the caller's permissions. The unpinned skill repository and global, noninteractive installation further expose users to unreviewed upstream changes; no actual compromise is demonstrated.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The semantic-markup installation command executes an unpinned npm CLI and selects a skill by name without an immutable repository revision. This is a supply-chain exposure rather than an MCP configuration issue, and a compromised upstream release could execute code or install malicious agent instructions.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The keyboard-a11y installation command uses a mutable npm CLI version and a mutable skill source. Although no MCP server is involved, following the command could execute compromised installer code or install unreviewed instructions globally.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The alt-text installation example allows npm to resolve a changing executable package and does not freeze the skill repository revision. This creates conditional supply-chain code-execution and agent-instruction risks, despite the benign accessibility purpose and inaccurate MCP label.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The a11y-audit installation example runs an unpinned npm CLI and installs a skill from a source not fixed to an immutable revision. This is not an MCP server reference, but compromised upstream content could affect the local environment or subsequent agent behavior.

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 · SKILL.md (reported line 44)May include surrounding context.

md
| 键盘可达与焦点管理 | [references/keyboard-focus.md](references/keyboard-focus.md) | 键盘, 焦点, focus, 陷阱, roving, tabindex, skip link |
| 读屏器实测方法 | [references/screen-reader-testing.md](references/screen-reader-testing.md) | 读屏, nvda, voiceover, 实测, live region, 测试脚本 |
| 审计与修复工作流 | [references/audit-remediation.md](references/audit-remediation.md) | 审计, 修复, 严重度, 报告, CI, 门禁, 流水线 |
| 语义标记(细粒度调用) | [skills/semantic-markup/SKILL.md](skills/semantic-markup/SKILL.md) | semantic-markup, 语义标签, dialog, 菜单, 标签页 |
| 键盘可达(细粒度调用) | [skills/keyboard-a11y/SKILL.md](skills/keyboard-a11y/SKILL.md) | keyboard-a11y, 模态, 焦点归还, 方向键 |
| 替代文本(细粒度调用) | [skills/alt-text/SKILL.md](skills/alt-text/SKILL.md) | alt-text, alt, 图标按钮, svg, 图表 |
| 无障碍审计(细粒度调用) | [skills/a11y-audit/SKILL.md](skills/a11y-audit/SKILL.md) | a11y-audit, 审计, 报告, 分级, 排期 |

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 44)May include surrounding context.

md
| 键盘可达与焦点管理 | [references/keyboard-focus.md](references/keyboard-focus.md) | 键盘, 焦点, focus, 陷阱, roving, tabindex, skip link |
| 读屏器实测方法 | [references/screen-reader-testing.md](references/screen-reader-testing.md) | 读屏, nvda, voiceover, 实测, live region, 测试脚本 |
| 审计与修复工作流 | [references/audit-remediation.md](references/audit-remediation.md) | 审计, 修复, 严重度, 报告, CI, 门禁, 流水线 |
| 语义标记(细粒度调用) | [skills/semantic-markup/SKILL.md](skills/semantic-markup/SKILL.md) | semantic-markup, 语义标签, dialog, 菜单, 标签页 |
| 键盘可达(细粒度调用) | [skills/keyboard-a11y/SKILL.md](skills/keyboard-a11y/SKILL.md) | keyboard-a11y, 模态, 焦点归还, 方向键 |
| 替代文本(细粒度调用) | [skills/alt-text/SKILL.md](skills/alt-text/SKILL.md) | alt-text, alt, 图标按钮, svg, 图表 |
| 无障碍审计(细粒度调用) | [skills/a11y-audit/SKILL.md](skills/a11y-audit/SKILL.md) | a11y-audit, 审计, 报告, 分级, 排期 |

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/audit-remediation.md (reported line 3)May include surrounding context.

md
# 审计与修复工作流(Playbook)

> 适用技能:[skills/a11y-audit](../skills/a11y-audit/SKILL.md)(流水线宿主:静态审计与报告产出)、[skills/alt-text](../skills/alt-text/SKILL.md)、[skills/keyboard-a11y](../skills/keyboard-a11y/SKILL.md)、[skills/semantic-markup](../skills/semantic-markup/SKILL.md)(被编排的修复执行技能)。

> 端到端地把"一团乱"变成"可交付的合规状态"。本篇是流程编排层:把其它 references、脚本、子技能串成一条流水线,并定义产出与完成标准。

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/semantic-html-aria.md (reported line 3)May include surrounding context.

md
# 语义化 HTML 与 ARIA(Playbook)

> 适用技能:[skills/semantic-markup](../skills/semantic-markup/SKILL.md)(语义标签选择、ARIA 用法与反模式修复)。

> 决定"什么时候用原生标签、什么时候上 ARIA、怎么上才不会更糟"。本篇是组件级可访问性的方法论核心。

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/semantic-markup/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/accessibility-engineer
  path: skills/semantic-markup/SKILL.md
  version: 1.0.0
  updated: 2026-09-02
  url: https://skills.sh/jakubkrehel/skills/better-accessibility

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 45)May include surrounding context.

md
| 读屏器实测方法 | [references/screen-reader-testing.md](references/screen-reader-testing.md) | 读屏, nvda, voiceover, 实测, live region, 测试脚本 |
| 审计与修复工作流 | [references/audit-remediation.md](references/audit-remediation.md) | 审计, 修复, 严重度, 报告, CI, 门禁, 流水线 |
| 语义标记(细粒度调用) | [skills/semantic-markup/SKILL.md](skills/semantic-markup/SKILL.md) | semantic-markup, 语义标签, dialog, 菜单, 标签页 |
| 键盘可达(细粒度调用) | [skills/keyboard-a11y/SKILL.md](skills/keyboard-a11y/SKILL.md) | keyboard-a11y, 模态, 焦点归还, 方向键 |
| 替代文本(细粒度调用) | [skills/alt-text/SKILL.md](skills/alt-text/SKILL.md) | alt-text, alt, 图标按钮, svg, 图表 |
| 无障碍审计(细粒度调用) | [skills/a11y-audit/SKILL.md](skills/a11y-audit/SKILL.md) | a11y-audit, 审计, 报告, 分级, 排期 |

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 45)May include surrounding context.

md
| 读屏器实测方法 | [references/screen-reader-testing.md](references/screen-reader-testing.md) | 读屏, nvda, voiceover, 实测, live region, 测试脚本 |
| 审计与修复工作流 | [references/audit-remediation.md](references/audit-remediation.md) | 审计, 修复, 严重度, 报告, CI, 门禁, 流水线 |
| 语义标记(细粒度调用) | [skills/semantic-markup/SKILL.md](skills/semantic-markup/SKILL.md) | semantic-markup, 语义标签, dialog, 菜单, 标签页 |
| 键盘可达(细粒度调用) | [skills/keyboard-a11y/SKILL.md](skills/keyboard-a11y/SKILL.md) | keyboard-a11y, 模态, 焦点归还, 方向键 |
| 替代文本(细粒度调用) | [skills/alt-text/SKILL.md](skills/alt-text/SKILL.md) | alt-text, alt, 图标按钮, svg, 图表 |
| 无障碍审计(细粒度调用) | [skills/a11y-audit/SKILL.md](skills/a11y-audit/SKILL.md) | a11y-audit, 审计, 报告, 分级, 排期 |

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/audit-remediation.md (reported line 3)May include surrounding context.

md
# 审计与修复工作流(Playbook)

> 适用技能:[skills/a11y-audit](../skills/a11y-audit/SKILL.md)(流水线宿主:静态审计与报告产出)、[skills/alt-text](../skills/alt-text/SKILL.md)、[skills/keyboard-a11y](../skills/keyboard-a11y/SKILL.md)、[skills/semantic-markup](../skills/semantic-markup/SKILL.md)(被编排的修复执行技能)。

> 端到端地把"一团乱"变成"可交付的合规状态"。本篇是流程编排层:把其它 references、脚本、子技能串成一条流水线,并定义产出与完成标准。

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/keyboard-focus.md (reported line 3)May include surrounding context.

md
# 键盘操作与焦点管理(Playbook)

> 适用技能:[skills/keyboard-a11y](../skills/keyboard-a11y/SKILL.md)(键盘可达基线、焦点可见与焦点管理落地)。

> 可操作性的地基。解决:所有功能能否不用鼠标完成、焦点可见且有序、模态/路由切换时焦点不丢失。

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/screen-reader-testing.md (reported line 3)May include surrounding context.

md
# 读屏器实测(Playbook)

> 适用技能:[skills/a11y-audit](../skills/a11y-audit/SKILL.md)(自动化审计后的实测验收环节)、[skills/keyboard-a11y](../skills/keyboard-a11y/SKILL.md)(焦点与播报类问题的读屏验证)。

> 自动化只能覆盖约 30% 的问题。本篇解决:为什么必须读屏实测、用什么工具、测什么、怎么写一份可复现的测试脚本。

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/keyboard-a11y/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/accessibility-engineer
  path: skills/keyboard-a11y/SKILL.md
  version: 1.0.0
  updated: 2026-09-02
  url: https://skills.sh/ibelick/ui-skills/fixing-accessibility

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 46)May include surrounding context.

md
| 审计与修复工作流 | [references/audit-remediation.md](references/audit-remediation.md) | 审计, 修复, 严重度, 报告, CI, 门禁, 流水线 |
| 语义标记(细粒度调用) | [skills/semantic-markup/SKILL.md](skills/semantic-markup/SKILL.md) | semantic-markup, 语义标签, dialog, 菜单, 标签页 |
| 键盘可达(细粒度调用) | [skills/keyboard-a11y/SKILL.md](skills/keyboard-a11y/SKILL.md) | keyboard-a11y, 模态, 焦点归还, 方向键 |
| 替代文本(细粒度调用) | [skills/alt-text/SKILL.md](skills/alt-text/SKILL.md) | alt-text, alt, 图标按钮, svg, 图表 |
| 无障碍审计(细粒度调用) | [skills/a11y-audit/SKILL.md](skills/a11y-audit/SKILL.md) | a11y-audit, 审计, 报告, 分级, 排期 |

> 路由规则:先判断任务属于「合规判据 / 语义标记 / 键盘焦点 / 读屏实测 / 审计修复」哪一类;做方法论决策读 `references/`,要落地某个具体能力直接调 `skills/`。

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 46)May include surrounding context.

md
| 审计与修复工作流 | [references/audit-remediation.md](references/audit-remediation.md) | 审计, 修复, 严重度, 报告, CI, 门禁, 流水线 |
| 语义标记(细粒度调用) | [skills/semantic-markup/SKILL.md](skills/semantic-markup/SKILL.md) | semantic-markup, 语义标签, dialog, 菜单, 标签页 |
| 键盘可达(细粒度调用) | [skills/keyboard-a11y/SKILL.md](skills/keyboard-a11y/SKILL.md) | keyboard-a11y, 模态, 焦点归还, 方向键 |
| 替代文本(细粒度调用) | [skills/alt-text/SKILL.md](skills/alt-text/SKILL.md) | alt-text, alt, 图标按钮, svg, 图表 |
| 无障碍审计(细粒度调用) | [skills/a11y-audit/SKILL.md](skills/a11y-audit/SKILL.md) | a11y-audit, 审计, 报告, 分级, 排期 |

> 路由规则:先判断任务属于「合规判据 / 语义标记 / 键盘焦点 / 读屏实测 / 审计修复」哪一类;做方法论决策读 `references/`,要落地某个具体能力直接调 `skills/`。

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/audit-remediation.md (reported line 3)May include surrounding context.

md
# 审计与修复工作流(Playbook)

> 适用技能:[skills/a11y-audit](../skills/a11y-audit/SKILL.md)(流水线宿主:静态审计与报告产出)、[skills/alt-text](../skills/alt-text/SKILL.md)、[skills/keyboard-a11y](../skills/keyboard-a11y/SKILL.md)、[skills/semantic-markup](../skills/semantic-markup/SKILL.md)(被编排的修复执行技能)。

> 端到端地把"一团乱"变成"可交付的合规状态"。本篇是流程编排层:把其它 references、脚本、子技能串成一条流水线,并定义产出与完成标准。

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/wcag-playbook.md (reported line 3)May include surrounding context.

md
# WCAG 2.2 合规模型(Playbook)

> 适用技能:[skills/alt-text](../skills/alt-text/SKILL.md)(SC 1.1.1 非文本内容判据落地)、[skills/a11y-audit](../skills/a11y-audit/SKILL.md)(合规审计的等级目标与 SC 依据)。

> 把"合规"从口号变成可执行的判定标准。本篇解决:该以哪个等级为目标、哪些成功准则(SC)最常踩雷、以及它们与法律的关系。

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/alt-text/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/accessibility-engineer
  path: skills/alt-text/SKILL.md
  version: 1.0.0
  updated: 2026-09-02
  url: https://skills.sh/addyosmani/web-quality-skills/accessibility

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 47)May include surrounding context.

md
| 语义标记(细粒度调用) | [skills/semantic-markup/SKILL.md](skills/semantic-markup/SKILL.md) | semantic-markup, 语义标签, dialog, 菜单, 标签页 |
| 键盘可达(细粒度调用) | [skills/keyboard-a11y/SKILL.md](skills/keyboard-a11y/SKILL.md) | keyboard-a11y, 模态, 焦点归还, 方向键 |
| 替代文本(细粒度调用) | [skills/alt-text/SKILL.md](skills/alt-text/SKILL.md) | alt-text, alt, 图标按钮, svg, 图表 |
| 无障碍审计(细粒度调用) | [skills/a11y-audit/SKILL.md](skills/a11y-audit/SKILL.md) | a11y-audit, 审计, 报告, 分级, 排期 |

> 路由规则:先判断任务属于「合规判据 / 语义标记 / 键盘焦点 / 读屏实测 / 审计修复」哪一类;做方法论决策读 `references/`,要落地某个具体能力直接调 `skills/`。

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 47)May include surrounding context.

md
| 语义标记(细粒度调用) | [skills/semantic-markup/SKILL.md](skills/semantic-markup/SKILL.md) | semantic-markup, 语义标签, dialog, 菜单, 标签页 |
| 键盘可达(细粒度调用) | [skills/keyboard-a11y/SKILL.md](skills/keyboard-a11y/SKILL.md) | keyboard-a11y, 模态, 焦点归还, 方向键 |
| 替代文本(细粒度调用) | [skills/alt-text/SKILL.md](skills/alt-text/SKILL.md) | alt-text, alt, 图标按钮, svg, 图表 |
| 无障碍审计(细粒度调用) | [skills/a11y-audit/SKILL.md](skills/a11y-audit/SKILL.md) | a11y-audit, 审计, 报告, 分级, 排期 |

> 路由规则:先判断任务属于「合规判据 / 语义标记 / 键盘焦点 / 读屏实测 / 审计修复」哪一类;做方法论决策读 `references/`,要落地某个具体能力直接调 `skills/`。

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/audit-remediation.md (reported line 3)May include surrounding context.

md
# 审计与修复工作流(Playbook)

> 适用技能:[skills/a11y-audit](../skills/a11y-audit/SKILL.md)(流水线宿主:静态审计与报告产出)、[skills/alt-text](../skills/alt-text/SKILL.md)、[skills/keyboard-a11y](../skills/keyboard-a11y/SKILL.md)、[skills/semantic-markup](../skills/semantic-markup/SKILL.md)(被编排的修复执行技能)。

> 端到端地把"一团乱"变成"可交付的合规状态"。本篇是流程编排层:把其它 references、脚本、子技能串成一条流水线,并定义产出与完成标准。