code-security-review

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection (LOW): The skill is designed to ingest and process untrusted code snippets, which creates a surface for indirect prompt injection if the analyzed code contains adversarial instructions.
  • Ingestion points: The workflow identifies user-provided source code, API parameters, and configuration files as primary inputs in SKILL.md.
  • Boundary markers: Absent. The instructions do not provide the agent with specific delimiters or commands to ignore instructions embedded within the code being reviewed.
  • Capability inventory: The skill focuses on generating text-based security reports and does not exhibit capabilities for file system modification, network exfiltration, or command execution.
  • Sanitization: Absent. There is no guidance on sanitizing or escaping the content of the code snippets before processing them into the final report.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:13 PM