lorenzini
Claude Code skills that wait for AI pull-request reviewers and judge whether their reports actually permit a merge, never inferring a pass.
Overview
lorenzini provides Claude Code skills that poll third-party pull-request reviewers, CodeRabbit, GitHub Copilot and Codex, and adjudicate whether their verdicts actually permit a merge. They do not review source code; they judge what the reviewer reported. The name comes from sharks' electroreceptors that find prey hidden in sand.
Its one rule is never infer a pass from absence. A verdict requires an explicit positive completion marker, otherwise polling continues; silence is not approval and an empty finding list does not mean no issues. The README documents a fail-open ledger of 11 incidents where earlier versions returned unearned passes, such as missing one of Copilot's two bot logins and declaring a PR with three actionable findings clean. Currently the CodeRabbit skill is live for every repository, while the Copilot and Codex skills are dormant after quota and subscription problems. Per the listing, Jev is used to judge whether verdicts permit a merge, and it works with Claude Code and Codex. Documentation is in English and Traditional Chinese.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this reviewer's report explicitly permit a merge?(inferred) | — |
| Yes / no | Has the reviewer finished (is there an explicit completion marker)?(inferred) | — |
How it works
- PR is opened and an AI reviewer is requested
- Skill polls for the reviewer's output
- Without an explicit completion marker, keep polling
- Judge whether reported findings permit a merge
- Record any fail-open incidents in the ledger
Strengths
- Fail-closed design
- Documented incident ledger
- Handles findings hidden in collapsed sections or threads
Caveats
- Depends on third-party reviewers' formats and quotas
- Only one reviewer skill currently live
- Jev's role not detailed in the README excerpt
At a glance
- Type
- Project
- Status
- beta
- Runtime
- Claude Code skills
- Has code
- yes
Use cases and integrations
How curated lists describe it
Claude Code skills that wait for CodeRabbit, Copilot or Codex to finish reviewing a pull request, then judge whether the verdict actually permits a merge.
MrJev/awesome-jev
Frequently asked questions
Does it review code?
No, it judges what the AI reviewer reported.
What is the core rule?
Never infer a pass from absence; require an explicit completion marker.
Which reviewers are supported?
CodeRabbit live; Copilot and Codex skills are dormant.
Related projects
Claude Code plugin with rules and checks that make agents prove their work; an optional Jev judge triages review findings.
PluginbetaMarkdown + shellPortable skill pack giving agents a conservative Jev second opinion on intent, completion, routing and bounded browser navigation.
ProjectbetaPythonLocal MCP review engine where each rule becomes a Jev YES/NO/N/A Choice over agent-supplied diff chunks, with located evidence.
MCP serverbetaTypeScriptDesktop app running coding agents in parallel git worktrees, with optional Jev judgments of each changed file against the task.
AppbetaLinear-backed work ledger for coding agents where a Jev-compatible model verifies done reports, blockers and plan changes against the task.
ProjectprototypeJavaScriptStop hook for Claude Code and Codex that refuses 'done' until a check passes after the last edit; Jev advises but never causes a block.
CLIbetaTypeScript