team-builder

Pass

Audited by Gen Agent Trust Hub on Mar 24, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads raw content from local markdown files to define sub-agent personas without proper isolation or sanitization.\n
  • Ingestion points: Agent persona instructions are read from markdown files in ./agents/ and ~/.claude/agents/.\n
  • Boundary markers: The skill lacks boundary markers or 'ignore embedded instructions' warnings when merging file content with the task description, using a simple concatenation: "{agent file content}\n\nTask: {task description}".\n
  • Capability inventory: The skill uses the Agent tool to spawn sub-agents which then execute logic based on the unvalidated file content.\n
  • Sanitization: No sanitization or validation is performed on the markdown files before they are used as instructions for the sub-agents.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 24, 2026, 07:43 AM
Security Audit — agent-trust-hub — team-builder