Back to skill

Security audit

AI新闻早报(中文版)

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a coherent AI news briefing generator, but it includes under-disclosed network routing and unsafe credential-handling guidance that users should review before installing.

Install only if you are comfortable giving the skill Tavily search access and, optionally, a SearXNG endpoint. Before using it, avoid the troubleshooting command that prints your API key, prefer HTTPS for any remote SearXNG instance, and remove or explicitly configure the hardcoded proxy in run_daily.sh unless you intentionally trust a service on 127.0.0.1:7897.

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 (3)

T09 · Insecure Skill Coding Practices

Warning
Location
INSTALL.md:123
Finding

Tavily API Key Exposed by Troubleshooting Command

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
INSTALL.md:110
Finding

Remote SearXNG Configuration Uses Unencrypted HTTP

Content
View full analysis
> ~/.openclaw/.env ``` `SKILL.md`: ```env TAVILY_API_KEY="tvly-xxxxxxxxxxxxxxxxxxxx" SEARXNG_BASE_URL="http://your-searxng-instance.com" ``` ```bash export SEARXNG_BASE_URL="http://your-searxng-instance.com" ``` ### Technical Analysis The documentation recommends an `http://` URL for a potentially remote SearXNG instance. HTTP provides neither transport encryption nor authenticated server identity. Consequently, network intermediaries can observe search terms and returned content. An active attacker in a position to modify network traffic can alter SearXNG responses, inject attacker-selected URLs or summaries, redirect requests, or impersonate the configured endpoint. This risk is particularly relevant because search results are treated as source material for an AI-generated executive briefing. Manipulated results could influence the agent's conclusions even without obtaining code execution. Plaintext HTTP may be acceptable for a service bound exclusively to loopback on the same host, but the documented hostname represents a remote deployment and does not state such a restriction. ### Attack Path 1. An operator copies the documented SearXNG configuration and replaces the placeholder with a remote host while retaining the `http://` scheme. 2. The Skill submits news queries over an unencrypted network connection. 3. An attacker controlling a network gateway, wireless access point, DNS path, or other intermediary intercepts the connection. 4. The attacker observes the queries or modifies the SearXNG response. 5. The agent consumes manipulated search results and may include false, misleading, or attacker-selected infor ...[truncated 598 chars]
Remediation
View remediation
> ~/.openclaw/.env ``` Apply the following controls: 1. Reject remote `http://` endpoints during configuration validation. 2. Permit plaintext HTTP only for explicitly documented loopback addresses such as `http://127.0.0.1`. 3. Require normal TLS certificate and hostname verification. 4. Do not provide an option that silently disables certificate verification. 5. Where practical, authenticate access to private SearXNG instances and store credentials in a secret manager. 6. Update every example in `INSTALL.md` and `SKILL.md` to use `https://`. 7. Consider validating response source domains and publication metadata before incorporating results into the briefing. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
run_daily.sh:6
Finding

Scheduled Script Forces Outbound Traffic Through an Undeclared Local Proxy

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (13)

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · INSTALL.md (reported line 32)May include surrounding context.

方法一:环境变量(推荐)

bash
# 编辑 ~/.openclaw/.env 文件
echo 'TAVILY_API_KEY="tvly-your-api-key-here"' >> ~/.openclaw/.env

# 重启 Gateway

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · INSTALL.md (reported line 33)May include surrounding context.

方法一:环境变量(推荐)

bash
# 编辑 ~/.openclaw/.env 文件
echo 'TAVILY_API_KEY="tvly-your-api-key-here"' >> ~/.openclaw/.env

# 重启 Gateway

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · INSTALL.md (reported line 110)May include surrounding context.

方法一:环境变量(推荐)

bash
# 编辑 ~/.openclaw/.env 文件
echo 'TAVILY_API_KEY="tvly-your-api-key-here"' >> ~/.openclaw/.env

# 重启 Gateway

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · INSTALL.md (reported line 123)May include surrounding context.

方法一:环境变量(推荐)

bash
# 编辑 ~/.openclaw/.env 文件
echo 'TAVILY_API_KEY="tvly-your-api-key-here"' >> ~/.openclaw/.env

# 重启 Gateway

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

存在描述与代码行为的不一致。描述声称该技能会使用 SearXNG 与 Tavily 两个引擎抓取情报,但脚本实际通过参数 --searxng_enabled=false --tavily_enabled=true 禁用了 SearXNG,因此与“多引擎(SearXNG 与 Tavily)”这一关键能力不符。此外,提供的代码片段只是一个外部调度脚本,主要作用是设置代理、生成日期、运行技能并记录日志;它并未直接体现去重、战略分级、洞察提炼和生成 5-9 条日报的实现。虽然这些功能可能存在于被调用的技能内部,但就该代码片段本身看,实际行为更接近“定时触发一个日报技能(且仅用 Tavily)”。因此应判定为不匹配。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The entire document specifies output formatting rules exclusively in Chinese and provides no indication that users may choose another language or locale. Under the stated policy, forcing a specific language without opt-in is a natural-language policy violation unless the locale restriction is explicitly documented and justified.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The installation guide instructs users to place API keys into configuration files and environment storage but provides no warning about credential sensitivity, file permissions, shell history exposure, or secret management practices. This increases the likelihood of accidental disclosure through copied configs, backups, logs, screenshots, or overly permissive local file access.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The documented manual triggers are very broad natural-language phrases such as '生成今天的AI日报' and '帮我搜一下今天AI圈有什么大事', which overlap with ordinary user requests. In agent platforms that auto-route or auto-invoke skills based on conversational intent, this can cause unintended activation, external web queries, and unplanned data flow to third-party services.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill name and all output instructions are framed as a Chinese AI daily briefing generator, and the required output templates hard-code Chinese titles and labels. There is no indication that the user can choose another language or that this locale restriction is explicitly justified as a region-specific tool.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The output section requires fixed Chinese strings such as "AI新闻早报", "速览", "洞察", and "今日研判" and says the agent must strictly use those templates. This enforces a specific language/locale in generated output without offering a choice or documenting why the restriction is necessary.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · run_daily.sh (reported line 11)May include surrounding context.

sh
DATE=$(date '+%Y-%m-%d')
LOG_DIR=~/.openclaw/logs
mkdir -p "$LOG_DIR"

# 调用 ai_daily_briefing_generator SKILL
cd ~/.openclaw/skills/ai_daily_briefing_generator

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

L13 的注释表明这里是在调用 ai_daily_briefing_generator 技能,而该技能清单描述其会调用多引擎(SearXNG 与 Tavily)抓取情报。但实际命令在 L19-L20 中将 searxng_enabled=false、tavily_enabled=true,运行行为与该技能所宣称的双引擎抓取意图不一致。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
76% confidence
Finding

The file presents the skill entirely in Chinese and the only example invocations are Chinese phrases, with no indication that other languages are supported or that Chinese is an optional locale. This can be a language/locale policy issue when a skill implicitly requires a specific language without user opt-in or justification.

Content

No source excerpt is available for this finding.