NYC

prompt-optimizer

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE] (SAFE): Comprehensive analysis of the 10 threat categories confirms no malicious patterns, command execution, or data exfiltration.
  • [NO_CODE] (SAFE): The skill consists entirely of documentation and methodology guides without any executable scripts or dependencies.
  • [INDIRECT_PROMPT_INJECTION] (SAFE): The skill ingests untrusted user requirements, representing a surface for indirect prompt injection. However, because the skill possesses no exploitable capabilities (no network, shell, or file access), the risk is negligible.
  • Ingestion points: User-provided vague requirements identified in the Step 1 workflow of SKILL.md.
  • Boundary markers: Absent in the final generated prompt template.
  • Capability inventory: No subprocesses, file writes, or network operations exist in any of the skill files.
  • Sanitization: Input is structurally transformed via the EARS methodology into normative specifications, which implicitly filters unstructured injection attempts.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 04:54 PM