laravel-specialist

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-defined requirements to generate application architecture and code, which introduces an attack surface for indirect injection if the source data is malicious.\n
  • Ingestion points: User-provided specifications during the requirement analysis phase in SKILL.md.\n
  • Boundary markers: Absent. The skill lacks explicit delimiters to separate user-provided content from the agent's instructional logic.\n
  • Capability inventory: The agent is instructed to run system commands via php artisan for migrations, routing, and testing across all reference files.\n
  • Sanitization: The skill explicitly mandates that the agent must not skip validation on user input and must avoid raw SQL queries to prevent injection attacks (SKILL.md).\n- [COMMAND_EXECUTION]: The skill utilizes standard Laravel Artisan commands for project management and verification.\n
  • Evidence: Execution of php artisan migrate:status, php artisan route:list, php artisan test, and php artisan queue:work are integrated into the core workflow and validation checkpoints (SKILL.md).\n- [DYNAMIC_EXECUTION]: The skill involves the generation and subsequent testing of PHP scripts based on user requirements.\n
  • Evidence: Workflow steps include implementing models and building features followed by immediate execution of php artisan test to verify the generated code (SKILL.md, references/testing.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 04:40 PM
Security Audit — agent-trust-hub — laravel-specialist