skills.rest
vanducng
vanducngCOMMUNITY
Explore all skills by this author →
Website𝕏@vanducngGitHub

auto-loop

Automate Claude Code sessions toward verifiable goals with two-vote completion gates.

5|Updated May 2, 2026
Search Tags:#verification#audit#autonomy#state-management#claude-code#goal-pursuit
One-click install
npx skills add https://github.com/vanducng/skills --skill auto-loop-vanducng
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: auto-loop
Source: https://github.com/vanducng/skills/tree/main/skills/auto-loop
Command: npx skills add https://github.com/vanducng/skills --skill auto-loop-vanducng

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires jq, git, claude, codex, ccusage, python3, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Automates driving a Claude Code session toward a verifiable goal until completion or a hard cap fires, removing the need for manual babysitting during long-running refactors, migrations, or test sweeps.

Core Features & Use Cases

  • In-session automation: hosts a Stop-hook-driven loop that feeds prompts until a two-vote gate opens or a hard cap triggers.
  • Two-vote completion gate: a user verifier and a fresh-context audit must both vote "achieved" to terminate the loop.
  • Safe caps and graceful drain: iterations, tokens, and wallclock are bounded to prevent runaway work.
  • Codex delegation option: optional --codex path hands the goal to Codex for specialized handling.
  • Forensics and state tracking: records iteration counts, verifications, audits, and diffs for postmortem.

Quick Start

Provide a goal and a verifier command to start the in-house loop and let it drive iterations until completion or a cap fires.

Frequently Asked Questions about auto-loop

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate long-running code refactors without manual babysitting?▼

You can automate long-running code refactors by hosting a Stop-hook-driven loop that feeds prompts to Claude Code until a two-vote completion gate opens or a hard cap triggers, removing the need for manual babysitting.

How does the two-vote completion gate verify autonomous coding goals?▼

The two-vote completion gate verifies autonomous coding goals by requiring both a user verifier command and a fresh-context audit to vote "achieved", ensuring safe and observable progress before terminating the loop.

What stops an autonomous Claude Code session from running away with tokens?▼

Strict caps on iterations, tokens, and wallclock time prevent an autonomous Claude Code session from running away, triggering a graceful drain to stop runaway work safely.

Do I need git and jq installed to run automated verification loops?▼

Yes, you need git, jq, python3, and claude installed in your environment to run automated verification loops, as these dependencies are required for state-management, forensics, and driving the session.

Can I delegate batch verification tasks to Codex instead of Claude Code?▼

Yes, you can delegate batch verification tasks to Codex instead of Claude Code by using the optional --codex path, which hands the goal to Codex for specialized handling during the autonomous loop.

Related Skills

Explore More Software Engineering Skills→
happy-nuthappy-nut
5

setup

Install viban dependencies and configure CLAUDE.md workflows via interactive interview.

Community
Advanced
aslaiiaslaii

gsd-forensics

Analyze git history and planning artifacts to diagnose GSD workflow failures.

Community
Advanced
JKHeadleyJKHeadley
75

autonomous

Enforce stateful task tracking and verified stop conditions for autonomous coding sessions.

Community
Advanced
LamseyahEliasLamseyahElias

claude-code

Orchestrate Claude Code to automate coding tasks and git workflows.

Community
Advanced