caveman
claudeTalk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
Five procedures for testing a result before you believe it (design confounds, a null for a derived statistic, result autopsy, a second method, accession-to-paper crosswalk) and a short playbook of rules for agents doing research work.
Plugin result-checks from marketplace agentic-research-toolkit in ChenHsieh/agentic-research-toolkit. Commands read from the repo's manifest — 20 days ago.
accession-paper-crosswalkLink a sequence-data accession to the paper that describes it, and a paper to the data it deposited - covering BioProject (PRJNA/PRJEB/PRJDB), SRA (SRP/SRX/SRR/SRS), BioSample (SAMN), and GEO (GSE) identifiers. Handles the alias expansion that makes or breaks the search, routes around NCBI's near-em
agent-playbookRules for an agent doing research work. Use at the start of any analysis, job submission, reproduction or delegation, and again before reporting a result, or when unsure whether a result is done, wrong, or not yet measured.
design-confound-auditDetermine which questions a dataset's design can actually answer, before any model is fit. Cross-tabulates the factor of interest against every nuisance factor to find perfect confounding (a factor fully nested inside another cannot be separated from it by any statistical method), cells below an inf
ecosystem-mapperMap research fields, funding landscapes, and innovation ecosystems as interactive network graphs. Use this skill whenever the user wants to understand: who funds what in a field, who does research on what, how money flows through an academic or industry sector, which organizations dominate a space,
ml-genomics-best-practicesA checklist-driven workflow for running reproducible, defensible ML experiments in genomics. Enforces anti-circularity, class balance audit, LR baselines, bootstrap CIs, permutation tests, proper feature leakage prevention, and standardized reporting. Invoke whenever building a classification or reg
result-autopsyAdversarially audit a result you have already produced, before a reviewer or a collaborator does it for you. Given a finding, a results file, or a claim in a draft, this executes the checks most likely to kill it - shared variables on both sides of a comparison, censoring that points the same way as
scientific-schematicsCreate publication-quality scientific workflow diagrams as interactive HTML files with one-click SVG export for Figma. Use this skill whenever the user wants to build or recreate a flowchart, decision tree, pipeline schematic, methodology diagram (CONSORT, PRISMA), biological pathway, or any scienti
second-opinion-concordanceRe-derive a set of calls with a methodologically independent second method and mine the disagreements, which are the actual product. Covers choosing a method that fails differently from the first (not a parameter variant of it), running both over the same entities, building a concordance table, and
statistic-nullGive a derived summary statistic its own null distribution before believing anything it says. For any scalar computed from data - concentration, peakiness, enrichment ratio, importance mass, divergence, specificity, clustering score - construct a null that preserves everything except the claimed str
tikz-figuresBuild publication-quality figures with TikZ/pgfplots - manuscript figures whose fonts must match the body text, pipeline and architecture schematics, or any figure that should regenerate from a data file instead of being redrawn by hand. Routes to a reusable template directory rather than writing pg
trait-gene-minerSystematic workflow for mining experimentally validated genes associated with a biological trait from literature and ontology databases, with an interactive HTML dashboard output linking to source databases. Use whenever a user wants to collect, curate, or systematically identify genes linked to a p
Talk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
Persistent memory for Claude Code: the next session starts where the last one left off.
29 skills · 1 MCP server · hooks
tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.
1 skill
Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications
5 skills · hooks
Cross-session memory and token savings for coding agents.
6 skills · 1 agent · 1 MCP server · hooks
Package any website or local web build into a lightweight desktop app with Pake CLI.
1 skill
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.