Back to skill

Security audit

基于艾宾浩斯遗忘曲线的英语四级词汇记忆系统

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent CET-4 study helper, but its reminder setup uses persistent cron jobs and outbound account messages without enough user control.

Review this skill carefully before installing. Its study features are ordinary, but do not let it modify crontab, run scripts from /root, or send messages from a default account unless you explicitly approve the exact command, recipient, schedule, and removal method. Prefer an app-scoped reminder mechanism instead of host-level cron.

Vulnerability Patterns
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • 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)

T06 · System Persistence

Error
Location
references/schedule-config.md:11
Finding

Persistent Scheduled Execution Through Root-Scoped Cron Jobs

Content
View full analysis

Vulnerability Details

File Location: references/schedule-config.md, lines 11–46
Vulnerability Type: Persistent scheduled-task registration
Risk Level: High

The setup instructions direct the agent to create recurring cron jobs that execute a script from a root-scoped directory:

bash
0 8 * * * /root/.openclaw/scripts/cet4-reminder.sh
0 21 * * * /root/.openclaw/scripts/cet4-reminder.sh

The instructions also use the following command to detect an existing persistent task:

bash
crontab -l | grep cet4

Technical Analysis

The scheduled reminders are related to the declared learning-reminder functionality, but direct modification of crontab creates execution that persists across agent sessions and system restarts. This exceeds the privileges needed for vocabulary learning and progress tracking, particularly because the configured executable is located under /root.

The document does not specify:

  • Explicit confirmation immediately before changing the user's crontab.
  • Use of an unprivileged service account.
  • Ownership and permission validation for the reminder script.
  • Integrity verification before each execution.
  • Safe and idempotent installation or removal procedures.
  • Validation that an existing matching entry belongs to this Skill.
  • Restrictions preventing duplicate or attacker-controlled cron entries.

The command crontab -l | grep cet4 is also an insufficient ownership check. Any unrelated or attacker-created entry containing cet4 could be mistaken for the expected task, while legitimate entries without that literal string could be missed.

Although the Skill discloses reminder scheduling as a feature, disclosure does not eliminate the security implications of persistent execution. The package itself does not contain the referenced reminder script, so its eventual content and integrity cannot be established from the audited project.

Attack Path

  1. A us ...[truncated 1564 chars]
Remediation
View remediation

Remediation Suggestions

  1. Do not modify system or user crontabs automatically. Prefer a user-scoped, application-native reminder facility.
  2. Require explicit, informed confirmation immediately before creating any persistent task, including the exact schedule, executable path, recipient, and account that will run it.
  3. Never use root for this functionality. Create reminders under a dedicated unprivileged account with only the permissions required to read learning schedules and send the intended notification.
  4. Store the reminder executable in a fixed, non-user-writable location. Validate its canonical path, owner, permissions, and integrity before registration and execution.
  5. Ship and audit the reminder implementation with the Skill rather than referencing an absent external script.
  6. Use a unique, structured task identifier instead of grep cet4, and compare the complete expected entry before adding or changing it.
  7. Make installation idempotent and prevent duplicate entries.
  8. Provide a documented removal command and automatically remove the task when reminders are disabled or the Skill is uninstalled.
  9. Validate the message recipient and avoid relying implicitly on a default messaging account.
  10. If cron remains necessary, write a narrowly scoped wrapper that uses absolute paths, a restricted environment, safe permissions, controlled logging, and no interpretation of untrusted schedule or recipient values as shell syntax.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
Findings (13)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

L03 列出的触发短语中,“开始背单词”“复习单词”“每日单词学习”等表述较为宽泛,且用“等”结尾,未穷举触发范围或提供排除条件。这会让技能在一般学习对话中更容易与普通表达重叠,增加误触发风险。

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill instructs the agent to write persistent user schedule data and set a recurring task, but it does not require an explicit informed-consent step describing those side effects before performing them. This can lead to unexpected data persistence and unsolicited reminders, which is especially risky when the skill auto-infers setup flow from a missing record rather than first clearly asking for permission.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

YARA rule 'network_reconnaissance': Network reconnaissance and scanning patterns [hacktools]

Medium
Category
YARA Match
Confidence
65% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · references/cet4-vocabulary.md (reported line 1349)May include surrounding context.

md
| 同伴 |
| female | /ˈfiːmeɪl/ | adj./n. | 女性的 |
| fence | /fens/ | n. | 栅栏 |
| ferry | /ˈferi/ | n. | 渡船 |
| festival | /ˈfestɪvl/ | n. | 节日 |
| fetch | /fetʃ/ | v. | 取来 |
| fever | /ˈfiːvər/ | n. | 发烧 |
| few | /fjuː/ | adj./n. | 很少的 |
| fiber | /ˈfaɪbər/ | n. | 纤维 |
| fiction | /ˈfɪkʃn/ | n. | 小说 |
| field | /fiːld/ | n. | 田地 |
| fierce | /fɪəs/ | adj. | 凶猛的 |
| fifteen | /ˌfɪfˈtiːn/ | num. | 十五 |
| fifth | /fɪfθ/ | num. | 第五 |
| fifty | /ˈfɪfti/ | num. | 五十 |
| fight | /faɪt/ | v./n. | 战斗 |
| figure | /ˈfɪɡər/ | n. | 数字;人物 |
| file | /faɪl/ | n. | 文件 |
| fill | /fɪl/ | v. | 填满 |
| film | /fɪlm/ | n. | 电影 |
| filter | /ˈfɪltər/ | n./v. | 过滤 |
| final | /ˈfaɪnl/ | adj./n. | 最后的 |
| finally | /ˈfaɪnəli/ | adv. | 最终 |
| finance | /ˈfaɪnæns/ | n. | 金融 |
| financial | /faɪˈnænʃl/ | adj. | 金融的 |
| find | /faɪnd/ | v. | 找�

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The setup flow tells the operator to write scheduling state and add cron entries without clearly informing the user that persistent system changes will occur. Hidden persistence reduces user awareness and can normalize silent modification of the host environment for a low-risk educational skill.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs creating host-level cron jobs and using a script located under /root, which grants persistent privileged execution unrelated to a normal vocabulary-learning workflow. In this context, persistence and root-scoped filesystem paths expand the blast radius if the skill is misused, modified, or parameterized unsafely.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The reminder template performs outbound messaging via a default account to an explicit target, introducing account-driven actions beyond simple local study assistance. If account or target values are wrong, reused, or attacker-influenced, the skill could send unintended messages or leak user-related activity.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The script sends proactive messages to a user account without any visible privacy or outbound-action warning in the skill description or setup steps. For an educational skill, silent outbound messaging creates privacy and consent risks because users may not expect account-based notifications triggered by background jobs.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
85% confidence
Finding

Checking and relying on crontab entries indicates persistent state outside the immediate session, which increases risk when used by a simple study assistant. In context, this persistence is more dangerous because it survives normal interaction boundaries and can continue triggering actions without ongoing user awareness.

Content

Scanner excerpt · references/schedule-config.md (reported line 43)May include surrounding context.

设置前检查用户是否已有定时任务:

bash
crontab -l | grep cet4

如果没有,则添加;如果有,确认时间是否与用户设置一致。

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

SQP-3 applies to all file types and includes language or locale policy violations. This markdown file delivers all headings, notes, and definitions in Chinese, which effectively fixes the locale without any opt-in or alternative language guidance.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The entire skill file is written in Chinese and does not indicate that the user can choose another language or that the skill is intentionally limited to a Chinese-speaking context. Under the policy for natural-language violations, forcing a specific language without user opt-in can be a locale-policy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The file's instructions and sample interaction are entirely in Chinese, including guidance to ask the user for setup information, with no indication that another language is supported or that Chinese is required for a specific regional purpose. That can constitute a language-policy issue when the skill effectively forces one language without user opt-in.

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 uses Chinese throughout and does not indicate that the skill or documentation supports other languages or allows user opt-in for locale. Under the stated policy, forcing a specific language without user choice can be a natural-language policy violation.

Content

No source excerpt is available for this finding.