autoresearch

Automate iterative task execution with modify, verify, and keep-or-discard decisions.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/bobpullie/DnT-WesangGoon --skill autoresearch-bobpullie
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: autoresearch
Source: https://github.com/bobpullie/DnT-WesangGoon/tree/main/.claude/skills/autoresearch
Command: npx skills add https://github.com/bobpullie/DnT-WesangGoon --skill autoresearch-bobpullie

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill enables autonomous, goal-directed iteration by repeatedly modifying, verifying, and deciding to keep or discard changes until a goal is achieved or iterations are exhausted. It applies Karpathy's autoresearch principles to any task, supports interactive planning, security audits, shipping, scenario generation, and other commands, and it can bound loops with Iterations: N inline config.

Core Features & Use Cases

  • Interactive Setup Gate: ensures required context is collected before running any loop.
  • Subcommands: /autoresearch, /autoresearch:plan, /autoresearch:security, /autoresearch:ship, /autoresearch:scenario, /autoresearch:predict, /autoresearch:learn, /autoresearch:reason.
  • Deterministic Iteration: Modify → Verify → Keep/Discard → Repeat with gating and logging.
  • Git Memory: reads past iterations from git history to guide next steps.
  • Bounded Iterations: specify Iterations: N to cap the loop.
  • Output & Logging: creates structured outputs and result logs for audit and improvement.

Quick Start

Start an autonomous autoresearch loop by supplying a Goal, Scope, Metric, Direction, Verify, and (optionally) Iterations to bound the run.

Frequently Asked Questions about autoresearch

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

FAQPage Schema
How do I set up autonomous iteration for a software engineering task?

Autonomous iteration requires supplying a Goal, Scope, Metric, Direction, Verify, and optionally Iterations to bound the run. An interactive setup gate ensures all required context is collected before the loop starts.

What is the modify-verify-keep-discard loop for autonomous task execution?

The modify-verify-keep-discard loop is a deterministic iteration cycle that repeatedly modifies code, verifies results, and decides to keep or discard changes until a goal is achieved or iterations are exhausted.

Can I bound autonomous loops to prevent infinite iterations?

Yes, you can bound autonomous loops by specifying Iterations: N in the inline configuration. This caps the maximum number of cycles to prevent unbounded execution during task automation.

How does autonomous iteration use git history for decision making?

Autonomous iteration uses git memory to read past iterations from git history. This contextual data guides the next steps in the loop by providing a structured log of previous modifications and outcomes.

Does autonomous task execution include security auditing capabilities?

Yes, autonomous task execution includes a dedicated security audit subcommand. This performs security checks during the iteration loop alongside planning, scenario generation, and shipping subcommands.

What are the limitations of using bounded iterations for complex tasks?

Bounded iterations limit execution to a specified number of cycles, meaning complex tasks may exhaust iterations before achieving the goal. If the verify step fails consistently, changes are discarded and the loop ends without completion.