arena

Pass

Audited by Gen Agent Trust Hub on Nov 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the omc (Oh My Claude) CLI and oh-my-claudecode:executor tool to manage parallel candidate runs and worktree isolation.
  • Evidence: Phase A specifies the use of OMC_TEAM_WORKTREE_MODE=detached omc team --no-decompose ... and dispatching tasks to an executor.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted external data in the form of a 'brief' and outputs from multiple parallel 'candidates' which are then scored and synthesized.
  • Ingestion points: The skill reads candidate artifacts and rationales in Phase B and Phase D.
  • Boundary markers: No explicit delimiter or 'ignore' instructions are specified for candidate artifact processing.
  • Capability inventory: The skill has the capability to execute shell commands (omc), write to the filesystem (git worktrees/scratch directories), and verify artifacts via the verify skill.
  • Sanitization: The instructions do not define sanitization or validation steps for the candidate-generated content before synthesis or verification.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 8, 2026, 07:21 PM