From autoresearch
This skill should be used when the user asks to "set up an experiment", "initialize a research run", "create experiment config", "prepare experiment branch", or needs to configure the autoresearch experiment environment before starting an autonomous optimization loop.
How this skill is triggered — by the user, by Claude, or both
Slash command
/autoresearch:setup-experimentThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Initialize the autoresearch experiment environment: create branch, config,
Initialize the autoresearch experiment environment: create branch, config, results file, and run baseline measurement.
Execute these steps in order before starting the experiment loop:
Propose a tag based on today's date and purpose (e.g., mar19-perf-opt).
Confirm the branch name experiment/<tag> does not already exist.
git checkout -b experiment/<tag>
Create experiment-config.yaml at the project root with all required fields:
tag: "<tag>"
target_files:
- "path/to/target.py"
eval_command: "just test" # or custom evaluation command
metric_name: "score"
metric_direction: "lower" # "lower" or "higher" is better
timeout_seconds: 120
results_file: "results.tsv"
baseline_description: "initial baseline"
Request each field value from the user if not obvious from context.
Create results file with header only:
printf 'commit\tmetric\tstatus\tdescription\n' > results.tsv
Add to .gitignore if not already present:
grep -q 'results.tsv' .gitignore 2>/dev/null || echo 'results.tsv' >> .gitignore
Read all target files and evaluation harness files to build full understanding of the codebase before experimenting. Identify:
Execute the evaluation command and record the baseline:
<eval_command> > run.log 2>&1
grep "^<metric_name>:" run.log
Record in results.tsv:
printf '%s\t%s\tkeep\tbaseline\n' "$(git rev-parse --short HEAD)" "<metric_value>" >> results.tsv
Present setup summary to the user:
Once confirmed, hand off to the research-loop skill or researcher agent.
Before proceeding, verify:
npx claudepluginhub hironow/dotfiles --plugin autoresearchSets up and runs autonomous experiment loops to optimize any target metric using git branches, autoresearch.md configs, bash benchmark scripts, and JSONL state logging. Activates on 'run autoresearch' or optimization loop requests.
Sets up autoresearch experiments interactively or via CLI for code optimization, collecting domain, target file, eval command, metric, direction, and evaluator.
Runs iterative experiments to optimize measurable metrics (speed, accuracy, config). Manages .lab/ directory for experiment history and autonomous workflow.