dev-orchestrator

Fail

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill uses imperative and restrictive language to override the agent's reasoning capabilities, stating "YOUR NEXT ACTION... MUST include a Bash tool call" and "There is no 'Option B'", which is a directive to bypass the AI's standard safety, manual review, and verification discretion.
  • [COMMAND_EXECUTION]: The instructions direct the agent to perform complex process management, including launching detached background tasks using tmux new-session -d and managing system environment variables such as env -u CLAUDECODE and AMPLIHACK_NONINTERACTIVE=1 to skip user-facing prompts.
  • [PRIVILEGE_ESCALATION]: The skill provides instructions for aggressive process control and termination, specifically recommending tmux list-sessions -F '#{pid}' | xargs -I{} kill {} to terminate background sessions, which can lead to system instability if used incorrectly by the agent.
  • [DATA_EXFILTRATION]: The orchestrator encourages automated reporting via gh issue create, instructing the agent to include "full error context, reproduction command, env details" which can inadvertently leak sensitive local environment variables, system paths, or project secrets to an external repository.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Nov 11, 2026, 03:30 PM