recipe-front-design

Pass

Audited by Gen Agent Trust Hub on Sep 23, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes an orchestration flow that ingests external data (user requirements via $ARGUMENTS) and passes it to multiple subagents without sufficient isolation or boundary markers.\n
  • Ingestion points: User requirements are processed in Step 1 and passed verbatim to codebase-analyzer (Step 2) and ui-analyzer (Step 3) as the governing source.\n
  • Boundary markers: The skill uses simple square-bracket placeholders (e.g., [confirmed requirements verbatim]) but does not instruct the agent to use distinct delimiters or tell subagents to ignore any instructions embedded within the requirements text.\n
  • Capability inventory: The skill invokes a suite of tools (dev-workflows-frontend namespace) capable of codebase analysis, UI spec generation, and technical design document creation, involving filesystem operations.\n
  • Sanitization: No sanitization or validation steps are defined to ensure the requirements do not contain malicious prompts intended to steer the behavior of the specialist agents.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 23, 2026, 10:09 AM
Security Audit — agent-trust-hub — recipe-front-design