debug-agent

Pass

Audited by Gen Agent Trust Hub on Feb 21, 2026

Risk Level: SAFE
Full Analysis
  • Indirect Prompt Injection (LOW): The agent processes untrusted inputs such as user-submitted bug reports and console logs.
  • Ingestion points: execution-protocol.md Step 1 (Error messages, user reports) and bug-report-template.md (Problem Description).
  • Boundary markers: No specific delimiters or markers are defined in the provided instruction files.
  • Capability inventory: The agent can search and read code via Serena MCP tools and write bug reports to the filesystem (.gemini/antigravity/brain/bugs/).
  • Sanitization: No explicit sanitization or filtering of external content is defined in the instructions.
  • Security Awareness (SAFE): The skill incorporates security best practices into its core workflow.
  • Evidence: The checklist.md requires the agent to verify that fixes do not introduce SQL injection, XSS, or authentication bypasses, and that sensitive data is not exposed in logs.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 21, 2026, 03:43 AM