skill-builder
Fail
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: CRITICALEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references a blog post at 'leehanchung.github.io/blogs/2025/10/26/claude-skills-deep-dive/' in reference.md. This URL has been flagged as malicious (URL:Blacklist) by automated security scanners. The presence of blacklisted URLs in documentation references constitutes a confirmed detection of risky content.
- [INDIRECT_PROMPT_INJECTION]: The skill serves as a meta-tool to generate other skills, agents, and commands from natural language descriptions, creating a significant attack surface for indirect prompt injection. 1. Ingestion points: User prompts defining the name, description, and functionality of new skills in SKILL.md. 2. Boundary markers: Absent. The generation process does not use delimiters or instructions to prevent the agent from obeying instructions embedded in the user's requirement description. 3. Capability inventory: The skill creates new files in the .claude/ directory and generates executable Python scripts (as shown in examples.md) which possess file-reading and analysis capabilities. 4. Sanitization: Absent. The skill does not validate or sanitize user input before interpolating it into generated markdown instructions or code.
- [DYNAMIC_EXECUTION]: The skill generates Python scripts (e.g., code_reviewer.py) and provides instructions for their installation and execution. Generating executable code from untrusted user prompts is a high-risk pattern that could lead to local code execution vulnerabilities.
Recommendations
- AI detected serious security threats
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata