skills.rest
ano4l
ano4lCOMMUNITY
Explore all skills by this author →
GitHub

run

Run a single autoresearch iteration by selecting an experiment, applying one change, and evaluating results.

Updated Jun 2, 2026
Search Tags:#git#workflow#evaluation#iteration#scripts#experimentation#autoresearch
One-click install
npx skills add https://github.com/ano4l/SiteRent --skill run-ano4l
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: run
Source: https://github.com/ano4l/SiteRent/tree/main/skills/autoresearch-agent/skills/run
Command: npx skills add https://github.com/ano4l/SiteRent --skill run-ano4l

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Run a single autoresearch iteration to guide selecting an experiment, apply one change, and evaluate results.

Core Features & Use Cases

  • Resolve experiment: If no experiment is specified, run a list and prompt the user to pick it.
  • Load context: Read the experiment config, strategy/constraints, results history, and checkout the experiment branch.
  • Decide what to try: Review results to determine the next change.
  • Make ONE change: Edit only the target file for this iteration.
  • Commit and evaluate: Commit the change and run the evaluation script.
  • Report result: Interpret the evaluation as KEEP, DISCARD, or CRASH.
  • Self-improvement check: After every 10 experiments, update the Strategy section with patterns learned.

Quick Start

Edit the target file to implement a single change, then run the evaluation to observe the outcome.

Frequently Asked Questions about run

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

FAQPage Schema
How do I automate running a single experiment iteration and evaluate the results?▼

Automate a single experiment iteration by selecting an experiment, applying one targeted change to a file, committing it, and invoking the evaluation script to classify the outcome as KEEP, DISCARD, or CRASH.

What is the best way to manage controlled code changes during iterative experimentation?▼

Manage controlled code changes by checking out a dedicated experiment branch, loading strategy and history context, making exactly one targeted edit, and committing it for auditable evaluation.

How does the evaluation workflow classify experiment outcomes after a change is committed?▼

The evaluation workflow classifies committed experiment outcomes by interpreting the evaluation script results, labeling each iteration as KEEP, DISCARD, or CRASH to guide the next decision.

Can I run iterative experiments without manually specifying which experiment to execute?▼

Yes, if no experiment is specified, the workflow lists available experiments and prompts the user to pick one before loading the config, strategy, and results history.

How does the experimentation workflow track and update long-term research strategy?▼

After every 10 experiments, the workflow automatically updates the Strategy section by extracting and recording patterns learned from the accumulated results history.

Does running a single experiment iteration require any external dependencies?▼

No external dependencies are required; the workflow operates using internal scripts to read experiment configs, checkout git branches, commit changes, and run evaluations sequentially.

Related Skills

Explore More Software Engineering Skills→
xlelibertyxleliberty

loop

Schedule iterative code execution and evaluation tasks for recurring research experiments.

Community
Advanced
xfs96192xfs96192

autoresearch

Automate iterative codebase improvements using Git history as memory.

Community
Advanced
TheRealSeberTheRealSeber
1

jade-core-atomic-commit

Stage and commit verified migration-rule changes as one atomic Git commit.

Community
Advanced
niznik-devniznik-dev
16

create-inspect-task

Create custom inspect-ai evaluation tasks with experiment-guided and standalone modes.

Community
Intermediate