brand-ambassador

Pass

Audited by Gen Agent Trust Hub on Apr 13, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill does not contain instructions to bypass safety guidelines, extract system prompts, or override agent constraints.
  • [DATA_EXFILTRATION]: No sensitive file paths, environment variables, or hardcoded credentials were found. There are no unauthorized network operations.
  • [REMOTE_CODE_EXECUTION]: The skill does not download or execute external code, scripts, or packages.
  • [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface by ingesting external data from web searches and user-provided brand information.
  • Ingestion points: User-supplied brand parameters and web search results for brand research (SKILL.md).
  • Boundary markers: Absent; there are no instructions to ignore embedded commands in the researched data.
  • Capability inventory: Limited to generating social media text and describing visual concepts; no shell or file-system access is requested.
  • Sanitization: No sanitization or validation steps are defined for the input data.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 13, 2026, 11:48 AM