first-pass
Claude Code plugin with rules and checks that make agents prove their work; an optional Jev judge triages review findings.
Overview
first-pass makes a coding agent check the code around a change, not just the lines it writes, and prove its work before calling anything done. Before coding, the agent answers ten questions against the real code, such as what happens if the change runs twice, stops halfway or an outside service times out. While building, anything the task did not ask for must earn its place or is listed as not built. The agent scopes its searches to the task, and before done a second agent reviews in a fresh context; done requires a test that failed before the change and passing CI in a clean checkout.
It is a Claude Code plugin; Cursor, Codex and other agents get the rules and skills but not the hooks. The decision-model part is optional and set up only when you ask: a Jev judge that the ship-check step consults on whether a review finding is real harm, which small findings to fix now, and what proof a small fix needs.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this review finding real harm? | yes · no |
| Choice | What proof does this small fix need? | — |
How it works
- Agent answers pre-change questions against the code
- Builds within scope and lists anything not built
- Second agent reviews; optional Jev judge triages findings
- Done only with a failing-then-passing test and clean CI
Strengths
- Evidence-based definition of done
- Jev judge optional
- Rules usable by several agents
Caveats
- Hooks are Claude Code only
- Jev judge requires a key
Getting started
Install the Claude Code plugin in the folder that holds your repos per the README; run the jev setup command only if you want the judge.At a glance
- Type
- Plugin
- Status
- beta
- Language
- Markdown + shell
- Runtime
- Claude Code plugin
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Claude Code plugin with rules and checks that make the agent look around a change before calling it done; an optional Jev judge decides whether a review finding is real harm and what proof a small fix needs.
MrJev/awesome-jev
Frequently asked questions
Is Jev required?
No, the Jev judge is optional.
What does the judge decide?
Whether a review finding is real harm, which small ones to fix now, and what proof a fix needs.
Which agents are supported?
Claude Code fully; Cursor, Codex and others get rules and skills.
Related projects
Desktop app running coding agents in parallel git worktrees, with optional Jev judgments of each changed file against the task.
AppbetaClaude Code skills that wait for AI pull-request reviewers and judge whether their reports actually permit a merge, never inferring a pass.
ProjectbetaStop hook for Claude Code and Codex that refuses 'done' until a check passes after the last edit; Jev advises but never causes a block.
CLIbetaTypeScriptSoftware-factory supervisor where Jev independently judges coding agents' completion, test sufficiency, progress and need for a human.
ProjectbetaPythonClaude Code plugin that scores a turn's redacted diff on eight Jev risk axes at Stop and asks Claude to reinvestigate once if high.
ProjectbetaGoGitHub Action where an AI agent turns an issue into a PR behind guards, checks, a reviewer and a Jev gate that escalates when unsure.
ProjectprototypeTypeScript