Back to skill

Security audit

CN Business Compliance Check

Security checks for vulnerabilities and agentic risk

Overview

This is a China-focused compliance checklist skill with no executable code; the main caveats are broad trigger wording and optional unpinned installs for related skills.

Install it as a China-specific compliance screening aid, not as a substitute for legal counsel. Be aware it may trigger on broad words like compliance, legal, or risk, and verify the two optional related skills separately before running the documented install commands.

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

T08 · Insecure Dependencies

Warning
Location
SKILL.md:113
Finding

Unverified Installation of Mutable Third-Party Skills

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 113
Vulnerability Type: Unverified third-party dependency installation
Risk Level: Medium

Complete Code Snippet:

markdown
安装:`clawhub install cn-ecommerce-compliance` / `clawhub install cn-advertising-compliance`

Technical Analysis

The documentation recommends installing two third-party Skills by package name without specifying immutable versions, cryptographic hashes, signatures, trusted publisher identities, or reviewed source repositories. Because neither dependency is included in this project, its contents and behavior were outside the auditable artifact.

Name-only installation resolves mutable registry content at installation time. If a package publisher or registry is compromised, a package name is taken over, or a future release becomes malicious, the installed component could contain attacker-controlled instructions, scripts, or configuration. Actual execution and privilege scope depend on the package manager, the retrieved package, and the permissions granted to the Agent environment.

Attack Path

  1. An attacker gains control of one of the referenced package names, compromises its publisher account, or compromises the package distribution channel.
  2. The attacker publishes a malicious or modified release under the expected name.
  3. A user or Agent follows the documented name-only installation command.
  4. The package manager retrieves the mutable attacker-controlled release because no version, digest, signature, or source identity is enforced.
  5. When the installed Skill is loaded or invoked, its malicious instructions or code operate with whatever tools and permissions the host grants it.

Impact Assessment

The vulnerable project itself contains no executable payload, credential access, persistence mechanism, or data-exfiltration logic. The potential impact originates from the unverified dependency and therefore cannot be ...[truncated 444 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the optional installation commands unless the additional Skills are necessary.
  2. Pin each dependency to an immutable, reviewed version or content digest rather than installing by name alone.
  3. Document and verify the expected publisher identity and authoritative source repository.
  4. Require package signature and checksum validation before installation.
  5. Review the complete dependency contents, including scripts and transitive dependencies, before approval.
  6. Use an allowlisted internal mirror or trusted registry and enable publisher-account protections.
  7. Install and run third-party Skills with least privilege in an isolated environment, denying filesystem, network, secret, and command-execution access unless explicitly required.
  8. Re-audit dependencies whenever their pinned versions are updated.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger list includes very broad everyday terms like “合规”, “合法”, “风险”, and “中国法律”, which can cause the skill to activate in many unrelated conversations. In an agent setting, unintended invocation can unnecessarily gate user requests through this skill, leading to workflow disruption, overblocking, and incorrect compliance framing for benign tasks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill is written as if Chinese law always applies, but it does not clearly state jurisdictional limits or require confirmation that the user’s activity, users, market, or entities are actually subject to PRC law. This can cause an agent to apply the wrong legal framework, producing misleading advice or unnecessary refusals when the request concerns another jurisdiction or mixed cross-border facts.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown file presents all instructions, classifications, and operational guidance only in Chinese. Under the language/locale policy criterion, forcing a specific language without user opt-in can be a natural-language policy violation unless the locale limitation is explicitly justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

This markdown file contains user-facing natural language content exclusively in Chinese and does not indicate that the user can select another language. Under the language/locale policy, forcing a specific language without opt-in can be a policy violation unless the regional constraint is clearly documented and justified.

Content

No source excerpt is available for this finding.