dzhaobio

Ycga Rms Creator — Data skill for Claude Code

Data community

Claude Code skill for authoring RMS ("Run My Samples") pipeline scripts on Yale McCleary/YCGA: grammar reference, manifest rules, per-step resource/staging decisions, reusable idioms, and a static val.

How to install Ycga Rms Creator

This entry records only its repository, not the path inside it, so there is no exact command to give. Open dzhaobio/ycga-rms-creator and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Ycga Rms Creator does

Alternatives in Data

Portaljs Add Resource — Add another file (resource) to an EXISTING dataset in a PortalJS portal — a data dictionary, methodology, or a 2.3k ★
Presentation Creator — Create presentations from project data 210 ★
Busabase — Open-source database & workspace for AI agents — structured data, durable knowledge, reusable skills, runnable 162 ★

README

ycga-rms-creator

A Claude Code [skill](https://code.claude.com/docs) for writing, extending, and reviewing `.rms` pipeline scripts for Yale YCGA's RMS ("Run My Samples") SLURM job-orchestration DSL on McCleary.

This is an *authoring* skill, not a *runbook*: it doesn't run a fixed sequence of operations for one recurring service (compare `ycga-rnaseq-deg`/`ycga-rnaseq-deg-as`, which do). Instead it's a grammar reference, a decision procedure for designing a new pipeline's steps, dependencies, and resource/staging choices, a set of proven idioms to reuse, and a linter to catch a known class of silent failure before handing a draft back.

Layout

SKILL.md -- the workflow: how to go from a task description to a correct .rms script, plus the must-know engine quirks and how to validate a draft before finishing.
references/ -- the detail:
- grammar.md -- the authoritative script grammar, read from the installed engine source (not just the docs, which have gaps and one directive that's flatly wrong for this build).
- manifest.md -- sample sheet / manifest format.
- resources_and_staging.md -- per-step ##ppn/##mem/##io/##local decision tree, with value tables from real production scripts.
- idioms.md -- reusable code snippets proven in production (PE/SE lane merge, genome-dispatch dict, cross-sample aggregation, etc.).
- engine_quirks.md -- directives that are parsed but functionally dead on this engine build, silent-failure traps, and every place the official docs disagree with what the installed engine actually does.
- cli_and_environment.md -- rms CLI flags, companion tools, live McCleary paths/queue names.
assets/ -- three short, clean, verbatim production scripts (two from James Knight, the RMS author; one from Dejian Zhao) as structural gold-reference examples.
scripts/validate_rms.py -- a static linter that mirrors the real parser's structural rules and flags the specific