brandkit
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (e.g., brand names, descriptions, and references) to generate image prompts and report summaries.
- Ingestion points: User-provided brand details and the kebab-case summary
<ten>used in filenames and report headers inSKILL.md. - Boundary markers: Not present. The skill interpolates user strings directly into markdown templates and file paths without instructions to ignore embedded malicious content.
- Capability inventory: The agent is instructed to create and modify files in the
llmwiki/directory (e.g.,llmwiki/wiki/draft/uiux/,llmwiki/wiki/index.md, andllmwiki/wiki/log.md). - Sanitization: No sanitization or validation is specified for the user-derived content before it is written to the filesystem.
- [COMMAND_EXECUTION]: The skill instructions direct the agent to perform file system operations, such as creating directories and writing markdown files. The use of user-supplied strings for filenames (
DDMMYY-<ten>.md) without explicit sanitization instructions could potentially lead to directory traversal if the underlying agent implementation does not enforce strict path sandboxing.
Audit Metadata