code-design
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to 'scan the codebase first' and 'read the codebase' to identify naming, structure, and error handling conventions. This ingestion of external data constitutes an attack surface where adversarial instructions hidden in the source code could attempt to manipulate the agent.
- Ingestion points: Instructions in
SKILL.mdunder 'Respect Existing Patterns' and 'Formatting' that involve reading external files. - Boundary markers: The skill does not specify any delimiters or safety markers to differentiate between code structure and potentially malicious embedded prompts.
- Capability inventory: The skill leverages file-reading capabilities and the execution of project-specific formatters (e.g.,
black,prettier,go fmt). - Sanitization: No validation or sanitization of the ingested code content is performed.
- [EXTERNAL_DOWNLOADS]: The
README.mdprovides an installation command that fetches resources from the author's public GitHub repository.
Audit Metadata