Skills
skills/redpanda77/skills/mission-control/Gen Agent Trust Hub

mission-control

Fail

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill generates an autonomous wrapper script that disables security guardrails.\n
  • Evidence: The template references/templates/autonomous/run-agent.sh executes claude -p --permission-mode bypassPermissions, which bypasses the agent's standard user confirmation requirements for all tool operations.\n- [REMOTE_CODE_EXECUTION]: The skill generates and executes several local shell scripts that govern the agent's behavior.\n
  • Evidence: The setup process in references/setup/autonomous.md and references/setup/human-in-the-loop.md creates multiple executable scripts including done-check.sh, run-agent.sh, and various validation scripts that are executed as subprocesses during the agent's work cycle.\n- [PROMPT_INJECTION]: The skill uses instructions to override default agent safety and interaction patterns.\n
  • Evidence: Templates such as references/templates/autonomous/run-agent.sh and references/templates/autonomous/prompts.md contain commands like 'Do not ask whether to continue', 'Stop only when done-check.sh passes', and 'Do not rely on TodoWrite or conversation memory for completion', which are designed to force the agent into an autonomous state and override platform-level safety constraints.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 14, 2026, 07:06 AM
Security Audit — agent-trust-hub — mission-control