multi-agent-workflow

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the oh-my-ag CLI and local shell scripts to manage sub-agents.
  • Evidence includes shell commands in SKILL.md and resources/examples.md such as oh-my-ag agent:spawn and .agent/skills/orchestrator/scripts/spawn-agent.sh.
  • The skill uses background processes (&) and synchronization (wait) to handle parallel agent execution.
  • [PROMPT_INJECTION]: The skill passes user-provided task descriptions directly into CLI arguments, which constitutes an indirect prompt injection surface.
  • Ingestion points: Task descriptions in the agent:spawn command arguments (e.g., SKILL.md Step 2).
  • Boundary markers: Double-quote delimiters are used for task strings, but no explicit instructions are provided to the sub-agents to ignore embedded prompt instructions.
  • Capability inventory: The skill is designed to execute shell commands and modify code using Serena MCP tools.
  • Sanitization: No explicit sanitization or validation logic is defined for the interpolated task descriptions.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 01:03 AM