Back to skill

Security audit

suilight

Security checks for vulnerabilities and agentic risk

Overview

The skill is a normal-looking SuiLight setup guide, but it asks users to clone, install, and run unpinned remote code that was not included in the reviewed artifact.

Install only if you are comfortable trusting the external SuiLight repository and its Python dependencies at install time. Prefer checking out a specific reviewed commit, using a virtual environment or container, reviewing requirements.txt before installation, and avoiding exposed secrets or sensitive directories in the runtime environment.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:29
Finding

Execution of Mutable Remote Code and Unverified Dependencies

Content
View full analysis
Remediation
View remediation
``` 3. Verify the checked-out commit or release artifact using a trusted signature or an out-of-band cryptographic digest. 4. Replace unconstrained dependencies with a lockfile containing exact versions and package hashes. Install with hash verification, for example: ```bash python -m pip install --require-hashes -r requirements.lock ``` 5. Review direct, Git-based, URL-based, editable, and transitive dependencies before installation. Avoid packages from untrusted indexes. 6. Use an isolated virtual environment or disposable container and run the application as a dedicated, non-root user. 7. Restrict filesystem access, mounted secrets, environment variables, and outbound network connectivity to the minimum required. 8. Add automated dependency vulnerability scanning, provenance verification, and reproducible-build checks to the release process. 9. Document the exact reviewed source revision and dependency-lock digest in `SKILL.md` so users can verify that they are executing the audited version. ]]>
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

L087 显示导入 KnowledgeSaloon,但 L089 却实例化 KnowledgeSalon(),两者拼写不同,说明文档对实际应使用的对象存在直接冲突。这会误导使用者理解技能提供的核心“跨域讨论”接口,属于文档与代码示例之间的意图不一致。

Content

No source excerpt is available for this finding.