grilling

Pass

Audited by Gen Agent Trust Hub on Aug 18, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies user-provided answers as the primary driver for reshaping the 'design tree' and determining the next 'frontier' of questions.
  • Ingestion points: User-provided answers during the interactive 'grilling' session described in SKILL.md.
  • Boundary markers: Absent; the instructions do not define delimiters or provide 'ignore embedded instructions' warnings for processing user data.
  • Capability inventory: The agent is explicitly authorized to access the filesystem, use local tools, and dispatch sub-agents to find facts based on the user-driven frontier (defined in SKILL.md).
  • Sanitization: Absent; there is no mention of validating, escaping, or filtering user input before it is used to determine which environment resources to access.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 18, 2026, 10:27 AM
Security Audit — agent-trust-hub — grilling