Endless Loop

    5

    Autonomous research and task loop that builds on previous findings to solve complex objectives while you sleep.

    $12

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    5 usesSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    endless-loop

    Example session with this skill installed

    /endless-loop "Find 5 free AI tools launched in March 2026" --max 3

    • Read your context and instructions
    • Compiled the endless-loop

    Running: /endless-loop "Find 5 free AI tools launched in March 2026" --max 3

    [Endless Loop] Iter 1 done | Found OpenClaw (100K+ GitHub stars), LTX 2.3 (4K video gen), GPT-5.4 free tier | 3 items found
    [Endless Loop] Iter 2 done | Found Gemini 3 Pro image gen, Wan 2.2 open-source video | 5 items found
    [Endless Loop] Iter 3 done | OBJECTIVE COMPLETE — 5 tools found with full details
    [Endless Loop] FINISHED — 3 iterations | Results saved to .claude/loop-state/results.md

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Autonomous Recursive Agent Loop

    Ditch the one-and-done prompting cycle. Endless Loop transforms your AI agent into an autonomous researcher that works continuously on complex objectives. Inspired by Karpathy's autoresearch, this skill enables an iterative workflow where each cycle builds upon the discoveries of the last, preventing hallucinations and repetitive outputs through a smart state-management system.

    What it does

    • Continuous Execution: Research, analyze, or optimize indefinitely (or up to a set limit) without manual intervention.
    • State-Safe Persistence: Every finding is logged to a local state directory. If the process crashes or you stop it, you can resume exactly where you left off.
    • Smart Context: Automatically builds a "memory" of previous iterations, including summaries and a "do-not-repeat" list, ensuring the agent always finds fresh information.
    • Karpathy Mode: Specifically designed for codebase optimization; the agent forms hypotheses, edits code, runs tests, and reverts regressions automatically.

    Why use this skill?

    Standard prompts often hit context limits or plateau in detail. This skill solves the "depth problem" by breaking a massive objective into incremental units of work. While you sleep, the agent can perform deep-dive market analysis, lead generation, or project-wide refactoring, delivering a comprehensive markdown report with timestamped findings and executive summaries.

    Supported Capabilities

    • Full integration with Web Search, Web Fetch, and File System tools.
    • Automated Bash execution for testing and environment checks.
    • Sub-agent orchestration for specialized sub-tasks.

    Best results with Claude Opus 4.6. Context window limits long runs to ~50 iterations per session. For runs longer than 3 hours, start a fresh session and the skill auto-resumes from saved state. Web research quality depends on Claude's built-in search capabilities.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    5 uses

    Times used to date

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 7 months ago

    Permissions

    Terminal / Shell
    Write Files
    Read Files
    Network Access

    File Scopes

    .claude/loop-state/**
    *.md

    Terminal used for sleep between iterations. Read/Write for state persistence and results files. Network for web research via Claude's built-in search. All files written to .claude/loop-state/ directory only.

    Listed7 months ago

    What's inside

    Frequently Asked Questions

    Unlike standard prompts that stop after one response, this skill creates a continuous loop where the agent remembers previous successes and failures, allowing it to dig deeper into research and solve complex tasks over several hours without manual input.

    The skill includes a smart state-management system that logs findings to a local directory; if you lose power or hit a rate limit, the agent can resume exactly where it left off without duplicating work.

    This skill is designed for agents with tool-calling capabilities and access to a local file system, specifically those using the Open Interpreter or similar frameworks that support bash execution and file manipulation.

    The agent manages its own context by summarizing past findings and maintaining a "do-not-repeat" list, which prevents it from hitting token limits or getting stuck in repetitive loops.

    Yes, you can define a maximum number of iterations or specific 'exit conditions' to ensure the agent stops once it has met your objective or reached a certain budget.

    You will receive the complete persistent state-management logic, the Karpathy-inspired optimization workflows, and comprehensive documentation on how to configure the agent for research or coding tasks.

    New in Agents & Orchestration