customize
Pass
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted data from existing project files and user interactions to customize rule templates, which constitutes a surface for indirect prompt injection.
- Ingestion points: Content is read from project files such as
package.jsonandCLAUDE.md, and gathered via theAskUserQuestiontool (SKILL.md). - Boundary markers: The skill does not provide instructions for the agent to use specific delimiters or protective instructions when interpolating gathered data into the target files.
- Capability inventory: The agent has permission to use
WriteandEdittools to modify filesystem content and a restrictedBashtool for search operations. - Sanitization: There is no defined process for validating or escaping the information collected before it is committed to the project rules.
Audit Metadata