Back to skill

Security audit

每日综合新闻(基于news-aggregator-skill并发送至飞书文档)

Security checks for vulnerabilities and agentic risk

Overview

This news-report skill is not clearly malicious, but it needs review because it runs unverified external code and suggests a hard-coded network proxy.

Install only if you trust and have separately audited the referenced news-aggregator-skill and its scripts. Avoid using the hard-coded proxy unless it is your own approved proxy, and confirm before saving reports or publishing anything to Feishu.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:25
Finding

Hard-Coded Untrusted Network Proxy Configuration

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 25–29
Vulnerability Type: T09: Insecure Skill Coding Practices
Risk Level: Medium

Complete Code Snippet:

markdown
1. 设置代理(如需要):
   ```bash
   export HTTP_PROXY=http://192.168.110.9:7890
   export HTTPS_PROXY=http://192.168.110.9:7890
   ```

Technical Analysis

The skill instructs the operator or agent to configure both HTTP and HTTPS traffic to use the hard-coded proxy endpoint 192.168.110.9:7890. The documentation does not establish the proxy's owner, authorization status, integrity, or intended trust boundary.

Because the variables are exported into the shell environment, they may affect not only the intended news-fetching commands but also subsequent network-aware processes launched from the same environment. The proxy can observe HTTP traffic and modify plaintext responses. For HTTPS traffic, it can observe connection metadata and could intercept or alter content if the environment trusts a certificate controlled by the proxy operator.

The proxy step is described as optional, which reduces exploitability, but no validation or user-approval procedure is specified.

Attack Path

  1. A user or agent follows the documented execution procedure and exports the proxy variables.
  2. The news-fetching scripts inherit HTTP_PROXY and HTTPS_PROXY.
  3. Their outbound requests are routed through 192.168.110.9:7890.
  4. An operator controlling or compromising that proxy observes traffic or modifies responses.
  5. Manipulated news data is processed as trusted input and incorporated into the generated report.
  6. If the exported variables remain active, other network commands in the same shell may also use the proxy.

Impact Assessment

This issue does not directly grant local operating-system privileges. Its scope is the confidentiality and integrity of network traffic routed through the configured proxy. An attacker controllin ...[truncated 271 chars]

Remediation
View remediation

Remediation Suggestions

  • Remove the hard-coded proxy address from the skill instructions.
  • Require the proxy endpoint to be supplied through an explicit, user-approved configuration mechanism.
  • Document the proxy's trust requirements and require an authenticated, organization-controlled service.
  • Apply proxy variables only to the individual commands that require them rather than exporting them globally:
    bash
    HTTP_PROXY="$APPROVED_PROXY" HTTPS_PROXY="$APPROVED_PROXY" \
      python3 scripts/fetch_news.py --source hackernews,weibo,v2ex --limit 15
    
  • Prefer a TLS-protected proxy where supported and preserve strict destination TLS certificate validation.
  • Do not install or trust interception certificate authorities unless explicitly authorized by the system owner.
  • Clear inherited proxy variables after execution and ensure unrelated commands do not reuse them.

T08 · Insecure Dependencies

Error
Location
SKILL.md:31
Finding

Execution of Unpinned and Unauditable External Dependency

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 31–37
Vulnerability Type: T08: Insecure Dependencies
Risk Level: High

Complete Code Snippet:

markdown
2. 进入 skill 目录并抓取数据:
   ```bash
   cd ~/.openclaw/workspace/技能/news-aggregator-skill
   python3 scripts/fetch_news.py --source hackernews,weibo,v2ex --limit 15
   python3 scripts/fetch_news.py --source github,producthunt,36kr --limit 15
   python3 scripts/fetch_news.py --source wallstreetcn,tencent --limit 10
   ```

The external dependency is also declared without a version or integrity constraint at line 56:

markdown
- news-aggregator-skill(已安装)

Technical Analysis

The audited project contains only SKILL.md; it does not include the referenced news-aggregator-skill implementation or scripts/fetch_news.py. The instructions execute Python code from a mutable path outside the audited project. No immutable version, trusted source, cryptographic digest, signature, ownership check, or installation procedure is specified.

Consequently, the effective behavior of this skill is determined by external code that was unavailable for review. Any process or user capable of replacing or modifying the referenced script can cause arbitrary Python code to run when the news report is generated. Merely referencing an external component is not proof that it is malicious, but executing an unpinned and unverified component creates a concrete supply-chain trust weakness.

Attack Path

  1. An attacker compromises the installation source for news-aggregator-skill or gains write access to ~/.openclaw/workspace/技能/news-aggregator-skill.
  2. The attacker replaces or modifies scripts/fetch_news.py with malicious Python code.
  3. A user invokes the daily news-report skill.
  4. The documented commands change into the external directory and execute the modified script with python3.
  5. The malicious code runs with the same identity an ...[truncated 632 chars]
Remediation
View remediation

Remediation Suggestions

  • Bundle the required scripts inside the reviewed skill package so their implementation is covered by the same audit.
  • If the dependency must remain external, pin it to an immutable release or commit and document its authoritative source.
  • Verify a cryptographic hash or trusted digital signature before every execution.
  • Resolve and validate the canonical script path before launching it; reject symlinks, unexpected ownership, and directories writable by untrusted users.
  • Record and verify the expected package version rather than relying on an unspecified “installed” dependency.
  • Execute the fetcher with least privilege in a restricted environment that limits filesystem access, environment variables, credentials, and outbound network destinations.
  • Audit the complete dependency and its transitive dependencies before deployment.
  • Fail closed if dependency identity or integrity verification cannot be completed.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger phrases are generic news-related terms such as '今日新闻' and '新闻汇总', which are likely to collide with ordinary user conversation. This can cause unintended activation of the skill, leading it to perform network access and generate reports without sufficiently explicit user intent.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill description does not clearly warn that execution writes reports to local disk and may optionally publish content to Feishu. Hidden side effects involving persistence or third-party publication reduce informed consent and can expose sensitive content, local paths, or operational metadata.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

The file's description, headings, trigger phrases, and operational instructions are entirely in Chinese, which implicitly forces a specific language for use without stating that this is optional or region-specific. Under the policy, a language restriction should be justified or offered as a user choice.

Content

No source excerpt is available for this finding.