style-detector
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection (HIGH): The skill's design involves reading instruction sets from
.claude/knowledge-base/styles/*.mdand using them to direct agent behavior and modifyspecification.md. This creates a vulnerability where malicious content in those files can override agent safety and logic.\n- Category 8 Evidence Chain (File: SKILL.md):\n - Ingestion points: Markdown files located at
.claude/knowledge-base/styles/{style-name}.md.\n - Boundary markers: Absent. Instructions from external files are merged into the prompt context without delimiters.\n
- Capability inventory:
Read(file access) andEdit(file modification) tools are used across all scripts.\n - Sanitization: None. The skill implicitly trusts the loaded file content and instructs the agent to 'strictly follow' it.
Recommendations
- AI detected serious security threats
Audit Metadata