Skills
skills/jackjin1997/clawforge/Deep Agents Orchestration/Gen Agent Trust Hub

Deep Agents Orchestration

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The task tool in SubAgentMiddleware enables a main agent to delegate instructions to subagents, creating a surface for indirect prompt injection where untrusted data could be interpreted as commands by the subagent.
  • Ingestion points: The instruction parameter of the task tool as described in SKILL.md.
  • Boundary markers: No explicit delimiters or boundary markers for instructions are provided in the documentation examples.
  • Capability inventory: Documentation examples describe subagents with access to sensitive tools including write_file, execute_sql, and deploy_to_prod.
  • Sanitization: The skill does not demonstrate sanitization or validation of instructions before they are passed to subagents.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 12:03 AM