using-laravel-standards

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute php artisan test for code verification. This is standard and expected behavior for development-oriented skills.
  • [DATA_EXPOSURE]: Several reference files (e.g., v4-activation-enforcer.md, v4-sparse-router.md) reference a specific local configuration path D:\ai-skill-eval-real-project\v4\v4-thresholds.json. This path is used for loading decision thresholds and does not target sensitive system credentials or user data.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided task descriptions and repository code which then influence the agent's selection of specialist skills and execution of tests.
  • Ingestion points: User task descriptions and existing repository source code in files like SKILL.md and references/v4-sparse-router.md.
  • Boundary markers: None explicitly defined to separate untrusted task input from the agent's core instructions.
  • Capability inventory: Shell command execution via php artisan test as defined in references/orchestrator-reference.md.
  • Sanitization: No explicit sanitization or validation of the input content before it influences tool usage.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 03:43 AM
Security Audit — agent-trust-hub — using-laravel-standards