security-patterns
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [Prompt Injection] (SAFE): No malicious override instructions or bypass attempts were found. The skill contains examples of insecure code, but they are clearly labeled as 'DANGEROUS' for educational purposes and are not intended to be followed as instructions by the agent.
- [Data Exposure & Exfiltration] (SAFE): The skill does not access sensitive files, environment variables, or hardcoded credentials. The permitted tools are restricted to local file reading.
- [Remote Code Execution] (SAFE): There are no external dependencies, remote script downloads, or execution patterns detected. All code examples are static documentation.
- [Indirect Prompt Injection] (SAFE): Although the skill is designed to analyze untrusted external code using 'Read' and 'Grep' tools, the risk is negligible as the skill lacks capabilities to perform network requests, execute shell commands, or write to the filesystem. (1) Ingestion points: Read, Grep, Glob tools. (2) Boundary markers: None. (3) Capability inventory: Read-only access to files via specified tools. (4) Sanitization: None.
- [Metadata Poisoning] (SAFE): The skill identifies itself as being based on an official Anthropic plugin. While this claim may be a branding exaggeration, the content itself is technically accurate and promotes security best practices rather than causing harm or misjudgment of safety.
Audit Metadata