resonance-backend

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (LOW): Indirect Prompt Injection surface. The skill ingests file content via read_file and possesses capabilities to execute commands and write files, creating a potential vector if malicious instructions are present in processed data.\n
  • Ingestion points: read_file is used for codebase analysis.\n
  • Boundary markers: Absent; no specific delimiters or instructions to ignore commands within file content are provided to the agent.\n
  • Capability inventory: run_command, write_file, edit_file, and read_file.\n
  • Sanitization: Absent; the skill focuses on validating the application's external inputs but does not sanitize the content it reads from the local environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:36 PM