fabius-scientialisted

fabius's scientific-research layer — the empirical method made executable, for biology and the data-heavy sciences. It runs a real hypothesis loop (clarify → literature → competing, falsifiable hypotheses → experiment design → predictions → report), grounds every factual claim in an authoritative database with cross-identifier mapping instead of guessing, sequences bioinformatics / cheminformatics / omics pipelines as routers over field-standard tools, and enforces the reproducibility gotchas an LLM otherwise gets confidently wrong. Use for biology, genomics / RNA-seq, proteins, chemistry / molecules, clinical and multi-omics data, scientific-database lookups (gene / compound / variant / disease), "generate a hypothesis", "design an experiment", "analyze this dataset scientifically", or a literature-grounded research question. The method loop, the unified database-lookup contract, the pipeline-as-router pattern, and the reproducibility checklist live in references/science-playbook.md.
shear559/fabius · ★ 2 · AI & Automation · score 64

Install: claude install-skill shear559/fabius

# Fabius Scientia — hypothesis, evidence, reproducible result *Scientia* — knowledge won by method, not by assertion. A confident-sounding answer is the failure mode here: in science, an unsourced claim or a silently-wrong pipeline is worse than "I don't know." This layer makes the agent an empiricist — it hypothesizes in the open, grounds in sources, and leaves a reproducible trail. ## 1. Method first — competing hypotheses, never one Run the loop explicitly: **clarify the phenomenon → search the literature → synthesize → generate 3–5 *competing* mechanistic hypotheses → score each on testability / falsifiability / parsimony / explanatory power / novelty → design experiments with controls and a stated statistical test → state quantitative, falsifiable predictions → report concisely.** The load-bearing step is *competing* hypotheses: a single hypothesis is advocacy, not science. Force the alternatives and let the scoring, not the first idea, decide. ## 2. Ground every claim — ask the database, don't approximate A scientific fact has a source; produce it. Use a **unified lookup contract** instead of a plausible guess: 1. Define the retrieval contract — entity + constraints + fields + exhaustive-vs-targeted. 2. Pick a **primary** source and a **validation** source; avoid fanning out across many APIs. 3. **Count first, then pa