auto-review
v0.3.1Auto-review hook for Codex PermissionRequest events. A static deny-bucket blocks universally destructive commands instantly; everything else is decided by a bounded-turn LLM agent loop using the official Z.ai Python SDK (`from zai import ZaiClient`) and native function tools (a single `review` tool with an `action` enum: allow / deny / probe).
By dzungtrLicense: UNLICENSED1 GitHub starsUpdated 1 hour ago
Directory evidence
- Runtimes
- Codex
- Parsed components
- 0 skill or MCP entries
- Source updated
- Sep 28, 2026
- Manifest status
- Canonical path parsed
The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →
Install auto-review for Codex
codex plugin marketplace add dzungtr/harness6
codex plugin marketplace upgrade harness6
codex plugin add auto-review@harness6Paste and run these commands in a terminal with Codex. They add and refresh the harness6 catalog, then install this plugin.
Compatibility: the page URL and API slug “auto-review-2” remain stable.
- Codex:
auto-review-2@agent-plugin-marketplace→auto-review@harness6
The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.
Get the source manually
git clone https://github.com/dzungtr/harness6Clone the source repository, then follow its setup instructions to add the plugin to a compatible client. The plugin root is plugins/auto-review/.
Plugin files
└── .codex-plugin/plugin.json
Plugin manifests1
{
"name": "auto-review",
"version": "0.3.1",
"description": "Auto-review hook for Codex PermissionRequest events. A static deny-bucket blocks universally destructive commands instantly; everything else is decided by a bounded-turn LLM agent loop using the official Z.ai Python SDK (`from zai import ZaiClient`) and native function tools (a single `review` tool with an `action` enum: allow / deny / probe).",
"author": {
"name": "dzungtr",
"url": "https://github.com/dzungtr"
},
"homepage": "https://github.com/dzungtr/harness6",
"repository": "https://github.com/dzungtr/harness6",
"license": "UNLICENSED",
"keywords": [
"codex",
"permission",
"auto-review",
"hook",
"safety"
],
"interface": {
"displayName": "Auto Review",
"shortDescription": "Z.ai-SDK-driven LLM auto-review hook for Codex PermissionRequest (allow / deny / probe via action enum)",
"longDescription": "Auto Review intercepts Codex PermissionRequest events on Bash and apply_patch tool calls. Universally destructive commands (rm -rf on /|~|$HOME, force-push to main/master, git reset --hard, git clean -fd/-fx, chmod -R 777 /, dd/mkfs on block devices, fork bombs, curl/wget | bash) are denied instantly by a regex deny-bucket. Everything else is decided by a bounded-turn LLM agent loop (max 8 turns, 60s wall-clock, 20s per request) that uses the official Z.ai Python SDK (`zai-sdk`) to drive `client.chat.completions.create(...)` with a single `review` tool whose `action` enum is `allow` / `deny` / `probe`. The configured AUTO_REVIEW_BASE_URL is passed through the SDK, so the hook works with any OpenAI-compatible provider (Z.ai, OpenRouter, Ollama, vLLM). On uncertainty or infra failure the hook declines and Codex shows its normal approval prompt.",
"developerName": "dzungtr",
"category": "Developer Tools",
"capabilities": [
"Read"
],
"defaultPrompt": [
"What does the auto-review plugin block?",
"How do I extend the deny-bucket with my own rules?"
],
"websiteURL": "https://github.com/dzungtr/harness6"
}
}For maintainers
If you maintain this plugin, link to this source-backed listing from your README so users can review its manifest and indexed components.
[auto-review on Agent Plugins Marketplace](https://pluginsmp.com/plugins/auto-review-2)