Agent Skills › alvinunreal/oh-my-opencode-slim › loop-engineering

loop-engineering

GitHub

实现 Loop Engineering 运行时，包含 Grill 编排器与 Monitor 监控器。支持多轮迭代执行、自动重试及人工审核回调，用于管理复杂 Agent 任务的循环执行流程。

src/skills/loop-engineering/SKILL.md alvinunreal/oh-my-opencode-slim

需要配置多轮迭代执行流程需要设置 Agent 任务的人工审核或失败升级机制初始化 Loop 引擎运行时环境

npx skills add alvinunreal/oh-my-opencode-slim --skill loop-engineering -g -y

More Options

Non-standard path

npx skills add https://github.com/alvinunreal/oh-my-opencode-slim/tree/master/src/skills/loop-engineering -g -y

Use without installing

npx skills use alvinunreal/oh-my-opencode-slim@loop-engineering

指定 Agent (Claude Code)

npx skills add alvinunreal/oh-my-opencode-slim --skill loop-engineering -a claude-code -g -y

安装 repo 全部 skill

npx skills add alvinunreal/oh-my-opencode-slim --all -g -y

预览 repo 内 skill

npx skills add alvinunreal/oh-my-opencode-slim --list

Frontmatter

{
    "name": "loop-engineering",
    "description": "Loop engineering runtime Grill + Monitor"
}

Loop Engineering Skill

Goal: "What are you trying to accomplish?"
Success criteria: "Describe how we know the loop succeeded."
Success type: choose from test, build, lint, command, fileExists, oracle, observer, manual. For CLI steps provide successCommand; for file detection provide successPath.
Execute agent: fixer / designer / explorer / librarian
Verify agent: oracle / observer / test
Max attempts (default 3)
Optional context files: which files or directories should be read before execution?

Listen to callbacks:
- onLoopComplete(loopID, success) → report final outcome
- onEscalated(loopID, reason) → escalate to human
- onManualReview(loopID, reason) → prompt human to approve/fail and call resolveManualReview(loopID, passed, reason)
Show current state and attempt count on each callback
For manual verification, present the failure reason before asking for pass/fail
If human forces cancellation, call cancel(loopID) through the orchestrator

Manual verification is the minimal on-ramp (autoresearch pattern). It pauses the loop until resolveManualReview is called. Do not auto-resolve.
BackgroundJobBoard signals (totalErrors, timeoutCount) are already baked into the runtime.