building-subagents

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface.
  • Ingestion points: The skill ingests untrusted data from the user-provided domain argument and outputs from WebSearch or WebFetch tools during the research phase (Phase 1).
  • Boundary markers: The provided agent template (Phase 2 and Agent Template section) does not utilize delimiters or specific instructions to the LLM to treat research-derived content as untrusted when drafting the subagent's guidelines or workflow.
  • Capability inventory: The skill allows for the creation of agents with powerful tools including Bash (command execution), Write/Edit (file system modification), and Task (agent orchestration).
  • Sanitization: There are no mechanisms described for sanitizing, escaping, or validating the external content before it is interpolated into the new agent's definition.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 05:33 PM