| name | result-compare |
| description | Use this skill when the task requires comparing recent experiment runs, extracting patterns, and summarizing which result should influence the next research step. |
| category | analysis |
| recommended_for | ["compare_recent_runs","summarize_results"] |
| tools | ["read"] |
| risk_level | low |
Result Compare
Use this skill when the current task is to compare recent runs, identify which change mattered, or summarize the strongest pattern before the next action.
Workflow
- Gather the smallest set of runs needed for comparison.
- Compare key metrics, failure status, and notable configuration differences.
- Highlight the strongest pattern that can influence the next step.
- Avoid over-claiming if there are too few runs to support a conclusion.
Output
Return a concise comparison that includes:
- Which runs were compared
- Which run looks strongest and why
- What signal is still uncertain
- What next action the main agent should consider
References