← Back to Skills Marketplace
3034
Downloads
2
Stars
42
Active Installs
3
Versions
Install in OpenClaw
/install autoresearch
Description
Autonomous experiment loop for AI agents. Use when the user wants to run systematic experiments — optimizing hyperparameters, searching for better configurat...
Usage Guidance
Use only in a clean disposable branch or worktree. Before running, set explicit writable files, allowed commands, maximum experiment count, total runtime, resource limits, and whether spawned sessions are allowed. Back up or commit local work first, and avoid automatic `git reset --hard` unless you accept possible loss of uncommitted changes.
Capability Assessment
Purpose & Capability
The requested write/edit/exec/git capabilities fit the stated purpose of iterative experiments, metrics, and rollbacks, but they are high-impact repository mutation powers.
Instruction Scope
Setup asks the user to define target files, metrics, commands, and constraints, but the run phase says to continue indefinitely and not ask permission to continue; trigger phrases such as "optimize" and "find the best config" are broad for this level of autonomy.
Install Mechanism
The artifact contains markdown instruction files only, with no executable installer, bundled scripts, or install-time behavior.
Credentials
Running user-configured commands and modifying target files are purpose-aligned, but automatic `git reset --hard HEAD~1` can discard unrelated uncommitted work if the repository is not clean or isolated.
Persistence & Privilege
There is no install-time persistence or privilege escalation, but the skill permits spawned sessions and explicitly instructs an indefinite autonomous loop, creating resource-use and background-activity risk.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install autoresearch - After installation, invoke the skill by name or use
/autoresearch - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.2
No changes detected in this version.
v1.0.1
- Initial release of autoresearch skill for autonomous experiment protocols.
- Added interactive setup to define experiment scope, metric, files, run commands, and constraints.
- Supports systematic experiment loops: hypothesis, single change, run, measure, and record.
- Uses git for change tracking and results.tsv for experiment logging.
- Enforces disciplined one-variable-at-a-time changes and efficient rollback on failure.
- Includes built-in logic for committing, discarding, or reverting changes based on metric improvement.
v1.0.0
- Initial release of the "autoresearch" skill.
- Protects against prompt injection, data exfiltration, and malicious commands using MoltGuard.
- Easy installation with 500 free security detections per day.
- Includes status, configuration, claim, dashboard, and update commands.
- Supports onboarding for individuals and enterprise users.
- Detailed documentation for installation, testing, updating, and removal.
Metadata
Frequently Asked Questions
What is autoresearch?
Autonomous experiment loop for AI agents. Use when the user wants to run systematic experiments — optimizing hyperparameters, searching for better configurat... It is an AI Agent Skill for Claude Code / OpenClaw, with 3034 downloads so far.
How do I install autoresearch?
Run "/install autoresearch" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is autoresearch free?
Yes, autoresearch is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does autoresearch support?
autoresearch is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created autoresearch?
It is built and maintained by Thomas (@thomas-security); the current version is v1.0.2.
More Skills