brainstorm-prompt-optimizer

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting untrusted user input (raw ideas) and interpolating them into a structured markdown template intended for downstream AI processes.
  • Ingestion points: Phase 2 parses the raw user idea for core actions and domains.
  • Boundary markers: The final output is encapsulated within fenced code blocks using a specific optimized-prompt label to separate it from other text.
  • Capability inventory: The skill uses Read and Bash tools solely to identify the project's tech stack and existing documentation (e.g., package.json, CLAUDE.md).
  • Sanitization: No programmatic sanitization is performed on the input idea; however, the output template explicitly includes security-focused sections like 'Negative Requirements' to ensure downstream generated specs address security constraints (e.g., preventing SQL injection or plain-text password storage).
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 04:23 PM
Security Audit — agent-trust-hub — brainstorm-prompt-optimizer