brand-strategy

Pass

Audited by Gen Agent Trust Hub on May 4, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-supplied data through the $ARGUMENTS variable and reads instructions from brand-voice.md without using boundary markers. This creates a surface for indirect prompt injection where external data could potentially override agent behavior. However, the risk is reduced because the skill lacks high-privilege capabilities.\n
  • Ingestion points: User arguments ($ARGUMENTS) and the brand-voice.md file.\n
  • Boundary markers: Absent; input is interpolated directly into the core logic instructions.\n
  • Capability inventory: No tools for network access, file system modification, or shell command execution are defined in the skill logic.\n
  • Sanitization: No validation or filtering is applied to ingested data.
Audit Metadata
Risk Level
SAFE
Analyzed
May 4, 2026, 02:11 PM
Security Audit — agent-trust-hub — brand-strategy