continuous-agent-loop
Solid具有质量门、评估和恢复控制的连续自主代理循环模式。
Install
Quality Score: 92/100
Skill Content
Details
- Author
- affaan-m
- Repository
- affaan-m/everything-claude-code
- Created
- 4 months ago
- Last Updated
- yesterday
- Language
- JavaScript
- License
- MIT
Integrates with
Similar Skills
Semantically similar based on skill content — not just same category
autonomous-loops
自主Claude代码循环的模式与架构——从简单的顺序管道到基于RFC的多智能体有向无环图系统。
autonomous-loop-patterns
This skill consolidates autonomous AI agent loop patterns: the Ralph Wiggum exit-code-2 pattern, our manage-loop (manage-continuation.json + manage-loop-hook.py), grind-loop.sh supervisor, completion detection strategies, safety caps, stall detection, and the tradeoffs between simple loops and sophisticated orchestration. Use when designing or debugging autonomous agent loops, choosing between loop approaches, or implementing safety limits. Do NOT use for choosing which command to run (use ai-coding-agents) or for task management (use linear).
autonomous-loop-patterns
Use when designing, reviewing, or debugging an autonomous AI agent loop: repeated agent execution, completion signals, checkpoints, supervisor respawn, stall detection, safety caps, and human handoff rules. Covers the core loop patterns from simple bounded runs through sentinel-based continuation, checkpoint-resume, and external supervisor loops. Do NOT use for choosing a specific agent product command (use agent-engineering or the product's docs), writing ordinary task instructions (use prompt-craft), or optimizing individual tool calls (use tool-call-strategy).
autonomous-loops
Patterns and architectures for autonomous Claude Code loops — from simple sequential pipelines to RFC-driven multi-agent DAG systems.
agent-loop
Detect requests for iterative autonomous agent loops and route to the appropriate loop executor