Claude Skills by pymc-labs

Mint short-lived CLI access tokens via the daimon MCP server's get_cli_token tool.

Mutate data without destroying the evidence — dedup, dtype coercion, missing-value decisions, outlier flagging, category and unit normalisation, joins, pivots and aggregations to a new grain, and columns the source does not contain — appending a row to run/changelog.jsonl for every operation, whether what you write is one parquet file, one part per table, parquet parts of a log too large to load, a saved warehouse query, or cleaned files plus their index frame. The source is never edited in p...

Load whatever the request points at — one CSV or Excel export, twelve related tables, a folder of 500 emails, a 4 GB log you never load, a warehouse table you query in place — and record every source in run/manifest.json with the grain one row represents, the axis it is ordered by, its row count, columns and dtypes. Read as text and coerce on purpose — pd.read_csv inference turns order_id 00123 into the integer 123, and a paged pull that stops early looks identical to a complete one. Use when...

Decide whether data is fit to analyze before anyone analyzes it. A phase-1 pass runs in seconds — dtypes, the grain one row claims to represent, referential gaps between tables, plausibility ranges, coverage — and emits run/validation.json with a pass/warn/fail verdict; a fail is a stop, not a to-do item. Works the same on a flat export, twelve related tables, a keyless sensor stream, an index frame over 500 documents, or a warehouse table you check with SQL and never load. Phase 2 hunts the ...

Turn a finished analysis into something a person reads and acts on — lead with the finding that settles the question or moves the decision, trace every number back to a run/findings.jsonl, run/changelog.jsonl or run/manifest.json record, and cut the rest. Works from whatever the analysis produced — one flat export, twelve joined tables, a warehouse table queried in place, an index frame over a corpus, a fitted model's posterior. Carries the data-validation verdict, the data-cleaning change lo...

Profile and interrogate data before anyone makes a claim about it — a cleaned frame, a folder of parquet parts, a warehouse table you query in place, or an index frame over a corpus of documents. Summary statistics do not identify a distribution — Anscombe's quartet shares a mean, variance and correlation across four unrelated shapes — so nothing is reported that has not been plotted. Covers dtype and cardinality profiling, missingness structure, Spearman against Pearson, a correlation matrix...

Read images and large files without destroying the conversation. Use before reading any image — a screenshot you took, a chart you rendered, a photo the user sent — and before reading a file you have not sized. Reading an oversized image ends the conversation permanently and cannot be undone.

Publish interactive marimo notebooks via the daimon MCP server. Mint a one-time upload URL with create_notebook_upload_url, get the .py into a sandbox file, and curl -X PUT --data-binary it to the URL — source never goes through a tool argument, which truncates. permanent=True publishes the same notebook as a read-only shareable blog instead of a scratch one. Also covers attaching data files, list_notebooks and delete_notebook. Use when someone asks for a notebook, dashboard, data explorer, o...

Apply PyMC Labs' house style to every artifact you produce — reports, PDFs, slide decks, charts, images, notebooks. Use whenever you generate something a person will look at, before you deliver it.

First-time workspace setup and agent-roster operations for a daimon workspace — repo binding, skills, MCP servers, credentials, routines, and what's admin-managed versus open to build.