codex-jev-preflight
Fail-open Codex UserPromptSubmit hook that asks Jev for task type, complexity, risk and execution mode and injects it as advisory context.
Overview
Codex Jev Preflight makes Codex's first decision about a task more consistent. Before Codex starts, a UserPromptSubmit hook asks Jev for the task_type, complexity, risk and execution_mode (for example inspect_then_act) and injects the validated result as advisory routing metadata in the context.
The assessment cannot override system or developer instructions or an explicit user request. It is fail-open: if Jev times out, hits a quota limit or returns an invalid answer, the task proceeds without it. Installation can be done by asking Codex to follow the README, with the key entered through hidden input rather than pasted into chat. There is a live demo page and a Chinese README.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What type of task is this? | e.g. code_change |
| Choice | How complex and how risky is it? | e.g. moderate · medium |
| Choice | Which execution mode fits? | e.g. inspect_then_act |
How it works
- User submits a prompt
- Hook asks Jev for task type, complexity, risk and mode
- Validate the answers
- Inject advisory metadata, or skip on failure
Strengths
- Consistent first-step routing
- Never blocks the task
- Key entered via hidden input
Caveats
- Advisory only; Codex may ignore it
Getting started
git clone https://github.com/wellkilo/codex-jev-preflight.git, run codex-jev-configure, then codex-jev-install.At a glance
- Type
- Hook
- Status
- beta
- Language
- Python
- Runtime
- Codex hook
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Fail-open Codex UserPromptSubmit hook that injects TypeSafe Jev pre-task routing metadata.
logicrw/awesome-jev-projectsFail-open Codex UserPromptSubmit hook that injects TypeSafe Jev pre-task routing metadata.
hellogumbo/awesome-jev
Frequently asked questions
Can it block a task?
No, it is fail-open and advisory.
What does it ask Jev?
Task type, complexity, risk and execution mode.
Can it override my instructions?
No, it cannot override system, developer or user instructions.
Related projects
Fork of codex-with-chatgpt where Jev plus fixed rules decide when to bring in web ChatGPT to plan or review and when Codex just acts.
ProjectbetaOne skill and MCP server for Claude Code, Codex and Hermes where Jev gates large-model turns, routes models and guards Git steps.
MCP serverbetaPythonMulti-agent Codex factory: Jev decides task shape, decomposition and routing, and failures are classified, not blindly retried.
ProjectprototypePythonClaude Code skill that splits work into PBIs for Codex, Antigravity or Claude sub-agents, using Jev for model choice and triage.
ProjectprototypePythonSoftware-factory supervisor where Jev independently judges coding agents' completion, test sufficiency, progress and need for a human.
ProjectbetaPythonOpenCode V2 server plugin that uses Jev to route requests between Plan, Build and custom agents and hand off unfinished work.
ProjectbetaTypeScript