from skills32
Human-guided, defensive exploratory data analysis for scientific datasets: capture context first, propose one step, run disposable scripts, and keep an append-o
This skill provides a guarded workflow for exploratory data analysis (EDA) on scientific datasets (CSV, FASTA, etc.). It enforces a context-first approach: capture the research question and constraints before loading data, create a single first-step analysis, run disposable PEP723-style scripts with uv run, and record actions and findings in an append-only session journal. The skill emphasizes safety, reproducibility, and small, reviewable steps rather than running large automated pipelines.
Use when a user supplies scientific data files and asks for guided exploration, summary statistics, or plots. Ideal for bio/chem/experimental data where domain context matters and exploratory actions must be human-approved. Also suited for sessions where auditability (journal, session folders) is required.
uv run; scripts are expected in a session scripts/ folder.Designed for agents that can manage files and run shell/python commands (OpenClaw-style agents, Claude Code, Codex). It assumes availability of uv tooling for isolated script execution.
This skill has not been reviewed by our automated audit pipeline yet.
