codex-orchestration-multi-model
Warn
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs users to install the
Codex-Orchestrationplugin from an unverified GitHub user repository (Cjbuilds/Codex-Orchestration). Since this source is not a trusted vendor, the installation and subsequent execution of this plugin represent an untrusted dependency risk. - [COMMAND_EXECUTION]: The provided Python integration examples utilize the
subprocess.runmodule to execute system-level CLI commands (e.g.,codex plugin list) and parse their output. While used here for status checks, this confirms a capability for local command execution within the skill's environment. - [INDIRECT_PROMPT_INJECTION]: The skill facilitates an orchestration workflow where a 'Planner' model generates content that is subsequently processed and reviewed by an 'Advisor' model.
- Ingestion points: The workflow initiates with a user-supplied task that the Planner role processes in
SKILL.md. - Boundary markers: There are no explicit instructions or delimiters shown to prevent the Advisor or Executor from following malicious instructions potentially embedded in the Planner's output.
- Capability inventory: The skill environment possesses the ability to write role configuration files to
.codex/agents/and execute shell commands via Python'ssubprocess. - Sanitization: No sanitization or validation logic is described for the data passed between the model roles.
Audit Metadata