Back to skill

Security audit

llm-wiki-expert

Security checks for vulnerabilities and agentic risk

Overview

The skill’s knowledge-base tools fit its purpose, with privacy considerations for optional cloud processing and saved queries.

Use a dedicated Wiki folder and keep credentials out of its sources and logs. Decide which questions and answers may be retained. Optional cloud embeddings send selected document and query text to a provider; use approved processing for confidential material or retain local search. Review and pin the installer and source revision before using the global, automatically confirmed installation examples.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (37)

Skill content is required to be written in Chinese without user opt-in.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The instruction mandates Chinese for skill body text, allowing English only for technical terms. It provides neither a user language choice nor a documented locale-specific justification.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

Although this is a CLI installer rather than an MCP server, unversioned npx skills can download and execute a changing npm package under the invoking user's privileges. The global, noninteractive installation also retrieves skill content without an immutable revision, creating a supply-chain exposure if either upstream is compromised.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

This example invokes an unpinned npm CLI, so following it can execute an upstream package version that has not been reviewed. The @wiki-blueprint selector identifies a subskill, not an immutable package or repository version; the MCP classification is inaccurate, but the supply-chain concern is valid.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The unversioned npx invocation can execute a changing upstream CLI, and the selected RAG skill is not tied to an immutable repository revision. This is an installation-time supply-chain risk rather than an MCP configuration vulnerability.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read, file_write, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

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
| 检索、查询、综合与升级 | `references/retrieval-and-synthesis-playbook.md` | grep full-text semantic retrieval query synthesis promote |
| 刷新、责任、反馈与治理 | `references/maintenance-governance-playbook.md` | owner freshness SLO audit feedback deprecate migration |
| 质量模型、测试集与门禁 | `references/quality-evaluation-playbook.md` | coverage deadlink orphan precision recall eval gate |
| 定义范围、受众、页面契约 | `skills/wiki-blueprint/SKILL.md` | blueprint charter scope audience schema naming |
| 摄取和归档原始来源 | `skills/wiki-source-ingestion/SKILL.md` | source ingest raw metadata immutable triage |
| 编译结构化知识页面 | `skills/wiki-knowledge-compilation/SKILL.md` | compile split merge concept entity summary crosslink |
| 进行有据查询与综合 | `skills/wiki-grounded-query/SKILL.md` | grounded query cite unknown confidence archive |

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/information-architecture-playbook.md (reported line 124)May include surrounding context.

md
本手册的「信息架构 / 页面模型」方法论由以下子技能在生成或重构 Wiki 结构时调用:

- [wiki-blueprint](../skills/wiki-blueprint/SKILL.md) — 新建或调整 Wiki 蓝图、页面边界与导航时,先读本手册 §1–§9。

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/wiki-lifecycle-playbook.md (reported line 123)May include surrounding context.

md
本手册的「Wiki 生命周期」操作契约覆盖以下子技能,运行时按阶段调取对应子技能:

- [wiki-blueprint](../skills/wiki-blueprint/SKILL.md) — Blueprint 阶段(§3 ### Blueprint)。
- [wiki-source-ingestion](../skills/wiki-source-ingestion/SKILL.md) — Ingest 阶段(§3 ### Ingest)。
- [wiki-knowledge-compilation](../skills/wiki-knowledge-compilation/SKILL.md) — Compile / Promote 阶段(§3 ### Compile / Promote)。
- [wiki-quality-audit](../skills/wiki-quality-audit/SKILL.md) — Audit 阶段(§3 ### Audit)。

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/wiki-blueprint/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/llm-wiki-expert
  path: skills/wiki-blueprint/SKILL.md
  version: 1.0.0
  updated: 2026-08-14
  url: https://skills.sh/daymade/claude-code-skills/llm-wiki-setup

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
| 刷新、责任、反馈与治理 | `references/maintenance-governance-playbook.md` | owner freshness SLO audit feedback deprecate migration |
| 质量模型、测试集与门禁 | `references/quality-evaluation-playbook.md` | coverage deadlink orphan precision recall eval gate |
| 定义范围、受众、页面契约 | `skills/wiki-blueprint/SKILL.md` | blueprint charter scope audience schema naming |
| 摄取和归档原始来源 | `skills/wiki-source-ingestion/SKILL.md` | source ingest raw metadata immutable triage |
| 编译结构化知识页面 | `skills/wiki-knowledge-compilation/SKILL.md` | compile split merge concept entity summary crosslink |
| 进行有据查询与综合 | `skills/wiki-grounded-query/SKILL.md` | grounded query cite unknown confidence archive |
| 增量更新与生命周期治理 | `skills/wiki-maintenance/SKILL.md` | refresh cascade stale conflict supersede audit log |

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/source-grounding-playbook.md (reported line 105)May include surrounding context.

md
本手册的「来源 / 证据 / 溯源」方法论由以下子技能在摄取与查询时调用:

- [wiki-source-ingestion](../skills/wiki-source-ingestion/SKILL.md) — 摄取格式、增量与重复判断、冲突表示时,先读本手册 §2–§6。
- [wiki-grounded-query](../skills/wiki-grounded-query/SKILL.md) — 证据层契约、来源优先级、精确事实工作流时,先读本手册 §1–§4。

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/wiki-lifecycle-playbook.md (reported line 124)May include surrounding context.

md
本手册的「Wiki 生命周期」操作契约覆盖以下子技能,运行时按阶段调取对应子技能:

- [wiki-blueprint](../skills/wiki-blueprint/SKILL.md) — Blueprint 阶段(§3 ### Blueprint)。
- [wiki-source-ingestion](../skills/wiki-source-ingestion/SKILL.md) — Ingest 阶段(§3 ### Ingest)。
- [wiki-knowledge-compilation](../skills/wiki-knowledge-compilation/SKILL.md) — Compile / Promote 阶段(§3 ### Compile / Promote)。
- [wiki-quality-audit](../skills/wiki-quality-audit/SKILL.md) — Audit 阶段(§3 ### Audit)。
- [wiki-grounded-query](../skills/wiki-grounded-query/SKILL.md) — Query 阶段(§3 ### Query)。

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/wiki-source-ingestion/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/llm-wiki-expert
  path: skills/wiki-source-ingestion/SKILL.md
  version: 1.0.0
  updated: 2026-08-14
  url: https://skills.sh/astro-han/karpathy-llm-wiki/karpathy-llm-wiki

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
| 质量模型、测试集与门禁 | `references/quality-evaluation-playbook.md` | coverage deadlink orphan precision recall eval gate |
| 定义范围、受众、页面契约 | `skills/wiki-blueprint/SKILL.md` | blueprint charter scope audience schema naming |
| 摄取和归档原始来源 | `skills/wiki-source-ingestion/SKILL.md` | source ingest raw metadata immutable triage |
| 编译结构化知识页面 | `skills/wiki-knowledge-compilation/SKILL.md` | compile split merge concept entity summary crosslink |
| 进行有据查询与综合 | `skills/wiki-grounded-query/SKILL.md` | grounded query cite unknown confidence archive |
| 增量更新与生命周期治理 | `skills/wiki-maintenance/SKILL.md` | refresh cascade stale conflict supersede audit log |
| 运行 Wiki 质量审计 | `skills/wiki-quality-audit/SKILL.md` | lint manifest link source index staleness evaluation |

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/retrieval-and-synthesis-playbook.md (reported line 129)May include surrounding context.

md
本手册的「检索阶梯 / 有据综合 / 结果晋升」方法论由以下子技能在查询与编译时调用:

- [wiki-grounded-query](../skills/wiki-grounded-query/SKILL.md) — 查询计划、答案证据标签、冲突与未知、语义检索时,先读本手册 §1–§9。
- [wiki-knowledge-compilation](../skills/wiki-knowledge-compilation/SKILL.md) — 综合结构、查询结果晋升为编译知识时,先读本手册 §4、§7、§9。

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/wiki-lifecycle-playbook.md (reported line 125)May include surrounding context.

md
- [wiki-blueprint](../skills/wiki-blueprint/SKILL.md) — Blueprint 阶段(§3 ### Blueprint)。
- [wiki-source-ingestion](../skills/wiki-source-ingestion/SKILL.md) — Ingest 阶段(§3 ### Ingest)。
- [wiki-knowledge-compilation](../skills/wiki-knowledge-compilation/SKILL.md) — Compile / Promote 阶段(§3 ### Compile / Promote)。
- [wiki-quality-audit](../skills/wiki-quality-audit/SKILL.md) — Audit 阶段(§3 ### Audit)。
- [wiki-grounded-query](../skills/wiki-grounded-query/SKILL.md) — Query 阶段(§3 ### Query)。
- [wiki-maintenance](../skills/wiki-maintenance/SKILL.md) — Refresh / Deprecate 阶段(§3 ### Refresh / Deprecate)。

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/wiki-knowledge-compilation/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/llm-wiki-expert
  path: skills/wiki-knowledge-compilation/SKILL.md
  version: 1.0.0
  updated: 2026-08-14
  url: https://skills.sh/astro-han/karpathy-llm-wiki/karpathy-llm-wiki

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
| 定义范围、受众、页面契约 | `skills/wiki-blueprint/SKILL.md` | blueprint charter scope audience schema naming |
| 摄取和归档原始来源 | `skills/wiki-source-ingestion/SKILL.md` | source ingest raw metadata immutable triage |
| 编译结构化知识页面 | `skills/wiki-knowledge-compilation/SKILL.md` | compile split merge concept entity summary crosslink |
| 进行有据查询与综合 | `skills/wiki-grounded-query/SKILL.md` | grounded query cite unknown confidence archive |
| 增量更新与生命周期治理 | `skills/wiki-maintenance/SKILL.md` | refresh cascade stale conflict supersede audit log |
| 运行 Wiki 质量审计 | `skills/wiki-quality-audit/SKILL.md` | lint manifest link source index staleness evaluation |
| 构建 RAG 检索管线 | `skills/rag-retrieval/SKILL.md` | rag retrieval embedding 向量检索 混合检索 rerank 重排 分块 chunk 检索评测 recall |

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/retrieval-and-synthesis-playbook.md (reported line 128)May include surrounding context.

md
本手册的「检索阶梯 / 有据综合 / 结果晋升」方法论由以下子技能在查询与编译时调用:

- [wiki-grounded-query](../skills/wiki-grounded-query/SKILL.md) — 查询计划、答案证据标签、冲突与未知、语义检索时,先读本手册 §1–§9。
- [wiki-knowledge-compilation](../skills/wiki-knowledge-compilation/SKILL.md) — 综合结构、查询结果晋升为编译知识时,先读本手册 §4、§7、§9。

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/source-grounding-playbook.md (reported line 106)May include surrounding context.

md
本手册的「来源 / 证据 / 溯源」方法论由以下子技能在摄取与查询时调用:

- [wiki-source-ingestion](../skills/wiki-source-ingestion/SKILL.md) — 摄取格式、增量与重复判断、冲突表示时,先读本手册 §2–§6。
- [wiki-grounded-query](../skills/wiki-grounded-query/SKILL.md) — 证据层契约、来源优先级、精确事实工作流时,先读本手册 §1–§4。

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/wiki-lifecycle-playbook.md (reported line 127)May include surrounding context.

md
- [wiki-source-ingestion](../skills/wiki-source-ingestion/SKILL.md) — Ingest 阶段(§3 ### Ingest)。
- [wiki-knowledge-compilation](../skills/wiki-knowledge-compilation/SKILL.md) — Compile / Promote 阶段(§3 ### Compile / Promote)。
- [wiki-quality-audit](../skills/wiki-quality-audit/SKILL.md) — Audit 阶段(§3 ### Audit)。
- [wiki-grounded-query](../skills/wiki-grounded-query/SKILL.md) — Query 阶段(§3 ### Query)。
- [wiki-maintenance](../skills/wiki-maintenance/SKILL.md) — Refresh / Deprecate 阶段(§3 ### Refresh / Deprecate)。

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/wiki-grounded-query/SKILL.md (reported line 7)May include surrounding context.

md
source:
  type: derived
  repo: skills-repo/llm-wiki-expert
  path: skills/wiki-grounded-query/SKILL.md
  version: 1.0.0
  updated: 2026-08-14
  url: https://skills.sh/astro-han/karpathy-llm-wiki/karpathy-llm-wiki

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 48)May include surrounding context.

md
| 摄取和归档原始来源 | `skills/wiki-source-ingestion/SKILL.md` | source ingest raw metadata immutable triage |
| 编译结构化知识页面 | `skills/wiki-knowledge-compilation/SKILL.md` | compile split merge concept entity summary crosslink |
| 进行有据查询与综合 | `skills/wiki-grounded-query/SKILL.md` | grounded query cite unknown confidence archive |
| 增量更新与生命周期治理 | `skills/wiki-maintenance/SKILL.md` | refresh cascade stale conflict supersede audit log |
| 运行 Wiki 质量审计 | `skills/wiki-quality-audit/SKILL.md` | lint manifest link source index staleness evaluation |
| 构建 RAG 检索管线 | `skills/rag-retrieval/SKILL.md` | rag retrieval embedding 向量检索 混合检索 rerank 重排 分块 chunk 检索评测 recall |
| RAG 检索深度实现(分块/重排/评测) | `references/rag-retrieval-playbook.md` | rag chunk embedding rerank recall 分块 重排 评测 RRF 混合检索 |

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/maintenance-governance-playbook.md (reported line 113)May include surrounding context.

md
本手册的「增量维护 / 治理」方法论由以下子技能在刷新与治理时调用:

- [wiki-maintenance](../skills/wiki-maintenance/SKILL.md) — 新鲜度 SLO、增量刷新、重命名/合并/拆分、变更与回滚时,先读本手册 §1–§10。

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/wiki-lifecycle-playbook.md (reported line 128)May include surrounding context.

md
- [wiki-knowledge-compilation](../skills/wiki-knowledge-compilation/SKILL.md) — Compile / Promote 阶段(§3 ### Compile / Promote)。
- [wiki-quality-audit](../skills/wiki-quality-audit/SKILL.md) — Audit 阶段(§3 ### Audit)。
- [wiki-grounded-query](../skills/wiki-grounded-query/SKILL.md) — Query 阶段(§3 ### Query)。
- [wiki-maintenance](../skills/wiki-maintenance/SKILL.md) — Refresh / Deprecate 阶段(§3 ### Refresh / Deprecate)。