Back to skill

Security audit

儿童成长跟踪与评估

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent child-growth tracking skill, but it persists sensitive records about a minor and includes hardcoded child-specific details without enough scoping or storage controls.

Install only if you understand this is designed around a specific child profile. Replace the hardcoded child name, birth month, and baseline data with your own private placeholders, use a private storage location, and require explicit confirmation before recording or reading child records. Avoid using it in shared workspaces or environments with broad backup/indexing access.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:65
Finding

Hardcoded and Persisted Sensitive Data About a Minor

Content
View full analysis

Vulnerability Details

File Locations:

  • SKILL.md:65-81
  • SKILL.md:293
  • references/benchmarks.md:5-16
  • references/template-monthly.md:7
  • references/template-quarterly.md:8

Vulnerability Type: Plaintext sensitive-data exposure and insecure data storage
Risk Level: Medium

Vulnerable Content

SKILL.md:65-81 defines a fixed plaintext workspace location for longitudinal records:

text
/workspace/projects/workspace/memory/xiyue/
├── README.md
├── profile.md
├── milestones.md
├── sports.md
├── subjects/
│   ├── math.md
│   ├── chinese.md
│   └── english.md
├── daily/
├── records/
│   ├── YYYY-MM-DD-月度速记.md
│   ├── YYYY-MM-季度评估.md
│   └── YYYY-MM-DD-事件记录.md
└── references/
    ├── 喜悦成长跟踪模型V2.md
    ├── 模板-月度速记.md
    └── 模板-季度评估.md

SKILL.md:293 embeds a specific child's birth month and age:

text
喜悦出生日期:2019年10月(2026年4月时6岁半)

references/benchmarks.md:5-16 embeds a detailed developmental baseline:

text
## 基线数据(2026年4月 · 6岁半)

| 维度 | 指标 | 基线值 | 备注 |
|:---|:---|:---|:---|
| D1 运动 | 跳绳/分钟 | 151个 | 仅训练4次 |
| D1 运动 | 街舞 | 7岁+提高班 | 学习5-6个月,连续升班 |
| D2 学业 | 识字量 | ~1000字 | 同龄平均200-500字 |
| D2 学业 | 数学 | 10以内加减法全掌握 | 含凑十法、相邻数 |
| D2 学业 | 逻辑题 | 坐标/分类/推理/找不同全对 | |
| D3 认知 | 游戏记忆 | 二周目反应速度超成年人 | |
| D3 认知 | 信息处理 | 明显优于同龄 | |
| D5 习惯 | 作息 | 规律 | |
| D5 习惯 | 学习习惯 | 能自主解题、按要求完成 | |

The templates also hardcode the child's name into generated record headings:

text
# 📝 喜悦月度速记 · {{date:YYMMDD}}
text
# 📊 喜悦阶段性评估 · {{date:YYMMDD}}

Technical Analysis

The distributed skill is represented as a reusable child-growth tracker, but it includes information associated with a specific minor: a name, birth month, age, physical performance, educational attainment, cognitive observations, habits, and activity history. These details are stored directly in package files rather than being supplied through a pri ...[truncated 2505 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the real child's name, birth month, age, and developmental measurements from the distributed package.
  2. Replace personal values with neutral placeholders such as CHILD_NAME, DATE_OF_BIRTH, and BASELINE_VALUE, or use unmistakably fictional sample data.
  3. Parameterize all template headings instead of embedding a child's name.
  4. Require explicit user confirmation before creating or updating persistent child records.
  5. Allow the user to choose a private storage location rather than imposing a predictable shared path.
  6. Apply least-privilege file permissions so only the intended user and authorized process can read the records.
  7. Encrypt sensitive records at rest when supported by the host environment, with keys stored separately from the records.
  8. Define retention and deletion controls, including an option to erase the complete profile and all generated records.
  9. Document which information is collected, why it is needed, where it is stored, and which tools or users may access it.
  10. Avoid collecting exact birth dates or other identifiers unless they are necessary; calculate age from user-supplied data without embedding that data in the reusable package.
  11. Add automated release checks that reject packages containing personal names, dates of birth, or populated child-development baselines.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (7)

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill declares broad everyday phrases such as child performance, emotions, relationships, and study progress as mandatory triggers, which can cause the skill to activate during ordinary conversation without clear user intent. In this context, unintended activation is especially sensitive because the skill is designed to collect, organize, and analyze detailed records about a child, increasing the risk of over-collection of personal and developmental data.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The skill metadata and operating instructions are entirely in Chinese and frame activation around Chinese trigger phrases only, with no indication that users may choose another language. Under the policy, forcing a specific language without opt-in can be a natural-language policy violation unless the locale restriction is explicitly justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The display name, description, and some tags are entirely in Chinese, which indicates a language-specific experience but provides no indication that users can choose another language or that the skill is intentionally limited to a Chinese-speaking region. This can violate language/locale policy when a skill imposes a specific language without explicit opt-in or documented justification.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest says the skill is based on an 8维发展模型 including D0关系, D1运动, D2学业, D3认知, D4社交, D5习惯, D6动机, and D7兴趣. This template instead implements sections only for D1-D6 and merges interest and motivation into a single 'D6 · 兴趣与内驱力', while omitting D0关系 and D7兴趣 as distinct dimensions, so the actual documented artifact does not match the claimed model.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

This markdown file contains only Chinese-language headings and guidance, and there is no natural-language indication that users may choose another language or that the skill is intentionally limited to a Chinese-speaking context. Under the policy rule for language/locale, forcing a specific language without user opt-in is a reportable issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The template’s visible tags, headings, and prompts are written entirely in Chinese, which imposes a specific language choice on users. Under the stated policy, a forced language or locale without an explicit opt-in or alternative can be a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

This markdown template presents all headings, labels, and instructions in a single language and does not indicate that Chinese is optional or that the skill is intended only for a Chinese-speaking audience. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.