sdrf-skills / rules
Cursor rule19 starsChanged 7 days agoMITView on GitHub
sdrf-skills — SDRF annotation workflows, setup, validation, ontology lookup for proteomics metadata
- Installs packages
We split this sdrf-skills.mdc into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
When working with SDRF files, proteomics metadata, or when the user asks to install dependencies or set up the environment, follow these …
- Distinctive: line 30
NEVER guess ontology accessions — always verify via OLS
- Very specific: line 31
NEVER invent SDRF column names — use the exact form in skills/sdrf-annotate/references/format-rules.md
- Distinctive: line 32
When a PXD accession is given, ALWAYS fetch project context + publication before annotating
- Distinctive: line 34
All ontology terms must include both label AND accession (e.g., "breast carcinoma" with EFO:0000305)
≈ “Common” is an estimate: at least 1 in 5 of the 6,350 Cursor rule files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
---
description: sdrf-skills — SDRF annotation workflows, setup, validation, ontology lookup for proteomics metadata
globs:
- "**/*.sdrf.tsv"
- "**/*.sdrf"
- "**/sdrf*"
alwaysApply: false
---
# sdrf-skills — SDRF Annotation Skills
Structured skills for expert-level SDRF (Sample and Data Relationship Format)
annotation in proteomics.
## What You Can Do
When working with SDRF files, proteomics metadata, or when the user asks to **install dependencies** or **set up the environment**, follow these workflows:
### Setup & Knowledge
### Core Workflows
- **Annotate a dataset**: `skills/sdrf-annotate/SKILL.md` — create an SDRF from a PXD and always have it independently reviewed; review an existing .sdrf.tsv; or plan. The only entry point for annotation and review
- **Plan metadata**: `skills/sdrf-annotate/SKILL.md` — pre-annotation strategy
- **Adversarial review**: `skills/sdrf-adversarial-review/SKILL.md` — require a fresh reviewer and hash-bound receipt
- **Contribute**: `skills/sdrf-contribute/SKILL.md` — PR to community repository
- **Campaign**: `skills/sdrf-campaign/SKILL.md` — screen a class of studies against your criteria, then loop annotate over the included ones
## Key Rules
1. NEVER guess ontology accessions — always verify via OLS
2. NEVER invent SDRF column names — use the exact form in `skills/sdrf-annotate/references/format-rules.md`
3. When a PXD accession is given, ALWAYS fetch project context + publication before annotating
4. Template selection should happen BEFORE annotation begins
5. All ontology terms must include both label AND accession (e.g., "breast carcinoma" with EFO:0000305)
6. Modification parameters must follow NT=;AC=;TA=;MT= format with correct UNIMOD accessions
## Bundled tools: contract and build
Install the helpers once: `pip install -e <sdrf-skills checkout>` (brings `sdrf-pipelines`), then `sdrf-tools doctor`.
Set `SDRF_SKILLS_ROOT` to the sdrf-skills checkout: the skills resolve `spec/`, `tools/` and `data/`
against it (Claude Code sets `CLAUDE_PLUGIN_ROOT` instead).
Two deterministic helpers keep annotation short and structurally valid. Run them as `sdrf-tools ...`:
- `sdrf-tools contract -t ms-proteomics [-t human ...]` — prints the column contract of the
template union: every column in order, required/optional/multiple, value form, permitted reserved
words and the ontologies to search. Read this instead of `TERMS.tsv` and the template YAMLs.
- `sdrf-tools build --samples samples.tsv --technical technical.tsv --files files.json
-t ms-proteomics [-t ...] -o output.sdrf.tsv` — expands a sample table (one row per source and
replicate; `files` = the fractions of one injection; `label` = `label free sample` or one channel)
plus a technical table (run-level `comment[...]` values, `|` between multiple values) into the SDRF:
fractions, technical replicates, channel rows, repeated columns and column order are decided by
code. It refuses, and writes nothing, on a file outside `files.json`, a file claimed twice, two rows
sharing a (source, biological replicate, technical replicate, fraction) coordinate, or an incomplete
channel map — it never fills a channel in.
Fix validation errors in the two tables and rebuild; never edit the SDRF by hand.
## Specification Data
The SDRF specification lives in the `spec/` git submodule:
- **Column definitions**: `spec/sdrf-proteomics/TERMS.tsv` — read for valid column names, ontology routing, allowed values
- **Template manifest**: `spec/sdrf-proteomics/sdrf-templates/templates.yaml` — read for template inventory and versions
- **Individual templates**: `spec/sdrf-proteomics/sdrf-templates/{name}/{version}/{name}.yaml`
## Important References
- Column-to-ontology mapping: the contract prints it per column (`<- NCBITaxon`, `<- UBERON, BTO`); TERMS.tsv `values` is its source
- Template layers and the selection decision tree: `skills/sdrf-annotate/references/templates.md`
- Common UNIMOD IDs: Acetyl=1, Phospho=21, Carbamidomethyl=4, Oxidation=35, TMT6plex=737
- Common error: UNIMOD:1 (Acetyl) ↔ UNIMOD:21 (Phospho) swap — most frequent mistake
Where files are headingNew Cursor rule files have more rules about planning and tracking tasks, and fewer about code style.
Of the Cursor rule files we read, the 3,978 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository6 other files it gives its agents.
- .opencode/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- sdrf-adversarial-reviewSkill
- sdrf-annotateSkill
- sdrf-campaignSkill
- sdrf-contributeSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.