Judge completed experiments and update wiki claims with verdicts.
npx skills add https://github.com/Lambenthan/empiricalwiki --skill exp-eval-lambenthan
Please help me install this Agent Skill. Skill: exp-eval Source: https://github.com/Lambenthan/empiricalwiki/tree/main/.claude/skills/exp-eval Command: npx skills add https://github.com/Lambenthan/empiricalwiki --skill exp-eval-lambenthan
It converts completed experimental results into rigorous, cross-checked wiki updates so your claim confidence, evidence, ideas, and graph relationships stay accurate without manual bookkeeping.
wiki/claims, wiki/ideas, and the originating wiki/experiments page (including the ## Claim updates section) while rebuilding derived graph artifacts.supports or invalidates edges into wiki/graph/edges.jsonl, then refreshes context_brief.md and open_questions.md to keep downstream reasoning consistent.Use case: you finish a batch of evaluation runs for a target hypothesis, and you want the knowledge graph to reflect which claims were validated, which were challenged, and what follow-up experiments are needed.
Run exp-eval on the completed experiment slug my-exp-001 in auto mode to update the related claim, idea status, and graph edges without pausing for confirmation.
High-intent search queries and answers about installing and using this skill.
To automatically update wiki claims after experiments, use a cross-model review process to judge results and map verdicts to confidence scores. This updates wiki pages and graph edges without manual bookkeeping.
Cross-model review for experiment evaluation uses a dedicated Review LLM to independently assess if results support, partially support, invalidate, or leave a claim inconclusive, ensuring rigorous and unbiased wiki updates.
You add supports or invalidates edges by mapping experiment verdicts to graph edges in edges.jsonl. The process also refreshes context_brief.md and open_questions.md to maintain consistent downstream reasoning.
Yes, you can update confidence scores and evidence synthesis automatically. The system evaluates completed experiments against prior evidence and produces deterministic wiki writes to keep claim status accurate.
Yes, experiment evaluation works with linked ideas and prior evidence. It reads experiments, claims, and ideas together, considering prior evidence from other experiments before updating the knowledge graph.
Automated wiki automation for scientific reviews requires completed experiments linked to target claims and depends on cross-model-review constraints. It produces deterministic verdicts but needs structured input formats to function correctly.
skyllwtReview research artifacts with scored verdicts and prioritized weaknesses.
StreamPilotOrgAnalyzes growth experiments and generates markdown reports with follow-up suggestions.
skyllwtCheck local screen sessions and remote endpoints for experiment status anomalies.
KITTranEvaluate completed experiments and update wiki knowledge graphs with verdicts.