laravel-review
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses dynamic context injection (the
!syntax) to execute localgit diffcommands. These commands are necessary to provide the agent with the code changes it is tasked to review. They are read-only and restricted to the repository context. - [PROMPT_INJECTION]: The skill processes untrusted code content, which presents a surface for indirect prompt injection. Ingestion points: Untrusted data enters via
git diffoutputs inSKILL.md. Boundary markers: Content is wrapped in code blocks, but no specific 'ignore instructions' warnings are used. Capability inventory: The skill usesgitand file reading tools (Read,Grep,Glob) as defined in the frontmatter. Sanitization: There is no automated sanitization of the code being reviewed. This risk is inherent to the tool's purpose and mitigated by the detailed audit checklist provided to the agent.
Audit Metadata