open-science-skills

Open Science Skills — 43 skills for experimental social science: research-repository scaffolding and auditing around a sources/ knowledge base (research-repo), session sitrep and closure for repo hygiene (sitrep, finished), decision-map research planning (research-wayfinder) and a single-sitting research interview for ideas, designs, and drafts (research-grill), conjoint design / cleaning / diagnostics, survey design, list experiments, cross-national design, live Qualtrics operations and pre-fielding audit (qualtrics-ops), fielded-data audit, topic modeling, LLM text-classification design and validation, a three-phase VLM-OCR pipeline (triage, evaluate, run, clean; plus an ALICE-cluster runbook, vlm-ocr-alice), a deliberative GPT-6 Astra and Claude Opus model committee chaired by a premier model that is never a member (Fable by default, Astra via /model-committee-astra), an advisor consult that escalates a working model's decision point to Fable, LLM calibration via token logprobs, citation checking, source-claim fact-checking, literature-review evidence mapping, figure and table design, house-style LaTeX formatting (paper-tex), Beamer slide-deck craft and QA (beamer), figure/table auditing, replication-package scaffolding and FAIR audit, hypothesis building, narrative building, pre-registration writing, methods reporting, creative divergence with an optional Codex mode (diverge), pre-submission review with an optional cross-model mode (paper-review-lite), presubmit, journal-review, author-side referee-response letters, a lead-detecting multi-model orchestration workflow that sends menial tool calls to Haiku and judgment work to Opus at a per-task effort (orchestrate), and the deliverable pipeline (deliverable-open, deliverable-intake, deliverable-lint: a manifest and append-only handoff per talk, module, paper, or chapter, dictated-braindump intake into a planning wiki, and a section-anchored whole-deliverable editorial lint run under a budget). All skills are avai

43 skills · 11 commands