codex-orchestrate
Warn
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command using a script located at '$HOME/projects/claude-plugins/agents-bridge/scripts/codex'. This script appears to be a local bridge to an external AI service.
- [COMMAND_EXECUTION]: The execution logic uses shell interpolation for a 'PROMPT' argument. The skill specifically warns that the agent must escape '$' characters, acknowledging that the command is vulnerable to shell expansion. If the prompt content (derived from user plans or code) contains shell-sensitive characters or malicious escape sequences, it could lead to arbitrary command execution on the host system.
- [DATA_EXFILTRATION]: Implementation details, interfaces, and code context are sent to the external 'codex' utility for processing. While this is the stated purpose of the skill, it involves transmitting potentially sensitive development context to an external bridge.
Audit Metadata