Resume Experiment
Resumes a paused or context-limited autoresearch experiment by checking out the experiment branch, reading the full history (config, strategy, results, git log), and reporting the current state before asking how to continue.
✨ What it does
- Lists experiments with status (active/paused/done) if none specified.
- Checks out the experiment branch and loads config, strategy, and full results history.
- Reads recent git log to understand recent changes.
- Summarizes current state including target, metric, experiment counts, best result, and recent patterns.
- Asks the user how to continue: single iteration, loop, or just review results.
- Hands off to /ar:run or /ar:loop with the experiment pre-selected when chosen.
🎯 When to use it
- When the user wants to pick up a previously started autoresearch experiment that was paused or interrupted.
- When the user runs the slash command /ar:ar-resume.
- When the user asks to continue an experiment but doesn't remember the details.
- When the user wants to review the status and results of an ongoing experiment before deciding next steps.
🚀 How to use
Trigger: /ar:ar-resume
Trigger with the slash command /ar:ar-resume, optionally specifying the experiment domain and name (e.g., /ar:ar-resume engineering/api-speed). If no experiment is given, it lists experiments and lets you pick. The skill then loads the full context and reports the state, then asks how you'd like to continue.
Example prompts:
/ar:ar-resume
/ar:ar-resume engineering/api-speed
📄 Output: A summary of the experiment's current state and a prompt for the next action.
📦 Add this skill to Claude Code
# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills
# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/engineering/autoresearch-agent/skills/ar-resume .claude/skills/ar-resumeSkill source: engineering/autoresearch-agent/skills/ar-resume/SKILL.md
⚠️ Good to know
Requires an existing autoresearch experiment with a branch and .autoresearch directory; the skill does not create new experiments.
❓ FAQ
What happens if I don't specify an experiment?
The skill lists all experiments with their status (active/paused/done) and lets you pick one.
What information does the skill load when resuming?
It checks out the experiment branch and reads the config.cfg, program.md, results.tsv, and recent git log.
What options do I have after the state is reported?
You can choose to run a single iteration (/ar:run), start a loop (/ar:loop), or just review the results.
🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.