heavy-brainstorm

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns, credential exposures, or dangerous command executions were identified within the skill instructions or reference files.
  • [PROMPT_INJECTION]: The skill uses persona-based instructions (e.g., "Think from this perspective ONLY", "not trying to be fair, balanced, or comprehensive") to guide the brainstorming process. These are standard reasoning techniques for the skill's stated purpose and do not attempt to override safety protocols or extract system prompts.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data (problems and constraints) through prompt interpolation.
  • Ingestion points: references/agent-prompt.md (variables {problem} and {constraints}).
  • Boundary markers: Absent; the inputs are interpolated directly into the persona prompt.
  • Capability inventory: The skill uses pure reasoning and does not specify any tools, network access, or file system operations.
  • Sanitization: Absent; however, the lack of side-effect capabilities means the risk is confined to the conversation context.
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 01:00 PM
Security Audit — agent-trust-hub — heavy-brainstorm