juvenal

分类: 工具与效率 | 上传者: zardus | 下载: 0 | 版本: v1.0（最新）

Create and run verified AI agent workflows using Juvenal

更新日志: Source: GitHub https://github.com/zardus/juvenal

目录结构

当前层级: plugin/skills/juvenal/

📄 SKILL.md 2.8 KB

SKILL.md

---
name: juvenal
description: Create and run verified AI agent workflows using Juvenal
argument-hint: "[goal or command]"
allowed-tools:
  - Bash
  - Read
  - Write
  - Glob
  - Edit
---

# Juvenal — Verified AI Agent Workflows

You are helping the user create and manage Juvenal workflows. Juvenal orchestrates AI coding agents through alternating implementation and verification phases, preventing agents from cheating on success criteria.

## What is Juvenal?

Juvenal is a framework where a non-agentic Python script orchestrates AI coding agents (Claude or Codex) through verified phases. Each phase has:
1. An **implementation prompt** — tells the agent what to build
2. One or more **checkers** — verify the work (scripts, agent reviewers, or both)

The key insight: the implementing agent and the checking agent are separate, so the implementer can't cheat by weakening tests.

## Workflow Formats

### 1. YAML (most expressive)

```yaml
name: "my-workflow"
backend: claude  # or "codex"
working_dir: "."
max_bounces: 999  # global bounce limit

phases:
  - id: setup
    prompt: "Set up the project scaffolding."
    checkers:
      - type: script
        run: "pytest tests/ -x"
      - type: agent
        role: tester

  - id: implement
    prompt_file: phases/implement/prompt.md
    bounce_target: setup  # on failure, bounce back to setup
    checkers:
      - type: script
        run: "make test"
      - type: agent
        role: senior-engineer
      - type: composite
        run: "pytest tests/ --tb=long"
        prompt: "Review test output:\n{script_output}"
```

### 2. Directory convention

```
my-workflow/
  phases/
    01-setup/
      prompt.md
      check-build.sh
    02-implement/
      prompt.md
      check-tests.sh     # script checker
      check-quality.md   # agent checker
```

### 3. Bare .md files

```
phases/
  01-setup.md       # gets default tester checker
  02-implement.md
```

## Checker Types

- **script** (`type: script`): Shell command, exit 0 = PASS
- **agent** (`type: agent`): AI agent that must emit `VERDICT: PASS` or `VERDICT: FAIL: reason`
- **composite** (`type: composite`): Script runs first, output fed to agent via `{script_output}`

## Built-in Roles

Agent checkers can use built-in roles: `tester`, `architect`, `pm`, `senior-tester`, `senior-engineer`

## CLI Commands

```bash
juvenal run workflow.yaml [--resume] [--backend claude|codex]
juvenal plan "goal description" [-o output.yaml]
juvenal do "goal description"
juvenal status
juvenal init [directory]
```

## Your Task

When the user invokes `/juvenal`, help them by:

1. If they provide a goal, create a `workflow.yaml` file for that goal
2. If they ask to run something, invoke `juvenal run` via Bash
3. If they need help, explain the workflow format

Always create workflows that are specific, testable, and have meaningful checkers.

登录后下载/点赞/收藏 ❤ 2 | ★ 0

juvenal

目录结构

SKILL.md

举报内容

提示