Evaluate completed experiment metrics and update linked idea statuses.
npx skills add https://github.com/duany049/multi-skill-orchestration --skill exp-eval
Please help me install this Agent Skill. Skill: exp-eval Source: https://github.com/duany049/multi-skill-orchestration/tree/main/.claude/skills/exp-eval Command: npx skills add https://github.com/duany049/multi-skill-orchestration --skill exp-eval
It turns completed experiment results into an evidence-based verdict that updates the linked idea’s lifecycle status and records why it succeeded or failed.
supports or invalidates edges (experiment → idea) and rebuilds derived graph artifacts (context brief and open questions).completed and requires linked_idea to be present.Use exp-eval to evaluate the completed experiment 'exp-slug-123' and update the linked idea with a single verdict: run exp-eval with argument-hint input 'exp-slug-123' and include --auto to apply wiki changes without pausing.
High-intent search queries and answers about installing and using this skill.
Evaluate completed wiki experiments by converting outcomes into support or failure verdicts for linked research ideas. The process reads experiment metrics and key results, invokes an impartial Review LLM, synthesizes a final outcome, and writes idea status transitions plus failure reasons directly to wiki pages.
Cross-model adjudication for experiment evaluation uses an impartial Review LLM to assess support versus refutation, then Claude synthesizes a final conservative outcome. This dual-model approach ensures rigorous evidence aggregation before updating the linked idea's lifecycle status.
Yes, you can aggregate sibling experiment evidence for a single research idea verdict. The evaluation process considers sibling experiments as aggregate evidence, applying their metrics and key results to determine a comprehensive support or failure outcome for the linked idea.
Experiment evaluation requires the wiki experiment to be marked as completed and a linked_idea field must be present. The evaluation refuses to run unless these safety constraints are met, ensuring only finished experiments with proper idea linkages are processed.
Automatically apply wiki updates after experiment evaluation by running the evaluation with the auto flag enabled. This writes changes to the linked idea page including status, failure reason, and date resolved, while adding supports or invalidates graph edges without pausing for confirmation.
During experiment evaluation, graph artifacts rebuilt include supports or invalidates edges connecting experiment to idea, plus reconstructed context brief and open questions. These derived graph artifacts ensure the research knowledge base reflects the new evidence-based verdict.
skyllwtDesign end-to-end experiment suites with ablation-driven iterative refinement for wiki-based research workflows.
zhuyingqinOrchestrate research idea discovery from direction to validated experiment plans.
sjtuytcOrganize papers, ideas, experiments, and claims into a structured knowledge base.
skyllwtExecute wiki-planned experiments by generating code, deploying, and collecting metrics.