Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Set up a new autoresearch experiment interactively. Collects domain, target file, eval command, metric, direction, and evaluator. Use when the user runs /ar:setup or asks to start optimizing a file with the autoresearch loop.
.claude/skills/alirezarezvani-setup/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -36% | 0% |
Set up a new autoresearch experiment with all required configuration.
/ar:setup # Interactive mode
/ar:setup engineering api-speed src/api.py "pytest bench.py" p50_ms lower
/ar:setup --list # Show existing experiments
/ar:setup --list-evaluators # Show available evaluatorsPass them directly to the setup script:
bashpython {skill_path}/scripts/setup_experiment.py \ --domain {domain} --name {name} \ --target {target} --eval "{eval_cmd}" \ --metric {metric} --direction {direction} \ [--evaluator {evaluator}] [--scope {scope}]
Collect each parameter one at a time:
Then run setup_experiment.py with the collected parameters.
bash# Show existing experiments python {skill_path}/scripts/setup_experiment.py --list # Show available evaluators python {skill_path}/scripts/setup_experiment.py --list-evaluators
| Name | Metric | Use Case | |------|--------|----------| | benchmark_speed | p50_ms (lower) | Function/API execution time | | benchmark_size | size_bytes (lower) | File, bundle, Docker image size | | test_pass_rate | pass_rate (higher) | Test suite pass percentage | | build_speed | build_seconds (lower) | Build/compile/Docker build time | | memory_usage | peak_mb (lower) | Peak memory during execution | | llm_judge_content | ctr_score (higher) | Headlines, titles, descriptions | | llm_judge_prompt | quality_score (higher) | System prompts, agent instructions | | llm_judge_copy | engagement_score (higher) | Social posts, ad copy, emails |
Report to the user:
/ar:run {domain}/{name} to start iterating, or /ar:loop {domain}/{name} for autonomous mode."| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 9,384 | 3,869 | -59% | 1 | 1 | 0% | 1,448 | 1,022 | -29% | 0 | 0 | — |
case-02 | fail→fail | 2,838 | 2,960 | +4% | 1 | 1 | 0% | 562 | 1,309 | +133% | 0 | 0 | — |
case-03 | fail→pass | 12,429 | 7,355 | -41% | 1 | 1 | 0% | 2,469 | 2,181 | -12% | 0 | 0 | — |
case-04 | fail→pass | 5,202 | 2,186 | -58% | 1 | 1 | 0% | 915 | 1,065 | +16% | 0 | 0 | — |
case-05 | fail→pass | 11,498 | 2,383 | -79% | 1 | 1 | 0% | 2,190 | 1,135 | -48% | 0 | 0 | — |
case-06 | fail→pass | 15,770 | 2,682 | -83% | 1 | 1 | 0% | 2,945 | 1,303 | -56% | 0 | 0 | — |
case-07 | fail→pass | 12,257 | 3,445 | -72% | 1 | 1 | 0% | 2,290 | 1,455 | -36% | 0 | 0 | — |
case-08 | fail→pass | 17,163 | 2,724 | -84% | 1 | 1 | 0% | 3,418 | 1,380 | -60% | 0 | 0 | — |
case-09 | fail→pass | 24,870 | 2,418 | -90% | 1 | 1 | 0% | 4,666 | 1,079 | -77% | 0 | 0 | — |
case-10 | fail→pass | 30,223 | 2,957 | -90% | 1 | 1 | 0% | 5,406 | 1,246 | -77% | 0 | 0 | — |
case-11 | fail→pass | 9,712 | 2,257 | -77% | 1 | 1 | 0% | 1,954 | 1,192 | -39% | 0 | 0 | — |
case-12 | fail→pass | 11,081 | 1,869 | -83% | 1 | 1 | 0% | 2,115 | 1,151 | -46% | 0 | 0 | — |
case-13 | fail→pass | 12,604 | 2,045 | -84% | 1 | 1 | 0% | 2,142 | 1,113 | -48% | 0 | 0 | — |
case-14 | fail→pass | 14,091 | 2,010 | -86% | 1 | 1 | 0% | 2,572 | 1,116 | -57% | 0 | 0 | — |
case-15 | fail→fail | 6,421 | 1,850 | -71% | 1 | 1 | 0% | 1,109 | 983 | -11% | 0 | 0 | — |
case-16 | fail→fail | 6,533 | 4,895 | -25% | 1 | 1 | 0% | 1,333 | 1,086 | -19% | 0 | 0 | — |
case-17 | fail→pass | 8,827 | 2,734 | -69% | 1 | 1 | 0% | 1,524 | 1,287 | -16% | 0 | 0 | — |
case-18 | fail→fail | 6,476 | 6,321 | -2% | 1 | 1 | 0% | 940 | 1,226 | +30% | 0 | 0 | — |
case-19 | pass→pass | 5,407 | 9,139 | +69% | 1 | 1 | 0% | 1,101 | 2,228 | +102% | 0 | 0 | — |
case-20 | pass→pass | 9,074 | 9,569 | +5% | 1 | 1 | 0% | 1,822 | 2,832 | +55% | 0 | 0 | — |
case-21 | pass→fail | 11,491 | 3,110 | -73% | 1 | 1 | 0% | 2,062 | 1,179 | -43% | 0 | 0 | — |
case-22 | fail→pass | 8,132 | 1,768 | -78% | 1 | 1 | 0% | 1,594 | 1,024 | -36% | 0 | 0 | — |
case-23 | fail→pass | 19,587 | 2,642 | -87% | 1 | 1 | 0% | 1,860 | 1,331 | -28% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 19 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +61 percentage points is the difference between those two pass rates over the 19 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.